Availability-Based Path Selection and Network Vulnerability Assessment

Size: px
Start display at page:

Download "Availability-Based Path Selection and Network Vulnerability Assessment"

Transcription

1 Avalablty-Based Path Selecton and Network Vulnerablty Assessment Song Yang, Stojan Trajanovsk and Fernando A. Kupers Delft Unversty of Technology, The Netherlands {S.Yang, S.Trajanovsk, Abstract In data-communcaton networks, network relablty s of great concern to network operators and customers, snce (1) the connecton requested by the customer should obey the agreed-upon avalablty, otherwse the network operator or servce provder may face lablty costs as stpulated n the Servce Level Agreement (SLA), and (2) t s mportant to determne the most vulnerable parts of a network to gan nsght nto where and how the network operator can ncrease the network relablty. In ths paper, we frst study the problem of establshng a connecton over at most k (partally) lnk-dsjont paths and for whch the avalablty s no less than δ (0 < δ 1). We analyze the complexty of ths problem n generc networks, Shared-Rsk Lnk Group (SRLG) networks and mult-layer networks. We subsequently propose a polynomal-tme heurstc algorthm and an exact Integer Non-Lnear Program (INLP) for avalabltybased path selecton. The proposed algorthms and two exstng heurstc algorthms are compared n terms of acceptance rato and runnng tme. Subsequently, n three aforementoned types of networks, we study the problem of fndng a (set of) network cut(s) for whch the falure probablty of ts lnks s bggest. Index Terms Avalablty, Routng, Survvablty, SRLG networks, Mult-layer networks, Mn-cut. I. INTRODUCTION Due to the mportance of data-communcaton networks, even short servce dsruptons may result n sgnfcant economc loss. Hence, survvablty mechansms to protect connectons are called for. For nstance, by allocatng a par of lnk-dsjont paths (nstead of only one unprotected path), data are transported by the prmary path, and upon lnk falure, can be swtched to the backup path. Ideally, a survvablty mechansm should also take nto account the relablty of lnks. For nstance, f both prmary and backup paths contan lnks that have a hgh probablty to become unavalable, then proper protecton cannot be provded. Connecton avalablty, a value between 0 and 1, s therefore mportant and refers to the probablty that a connecton (ncludng ts survvablty mechansm) s n the operatng state durng the requested lfe-tme of the connecton. However, a survvablty mechansm that does not allow for more than 2 lnk-dsjont paths for each connecton may stll fal to satsfy the customer s avalablty requrement and k > 2 lnk-dsjont paths may be needed. Obvously, the bgger k s, the greater the avalablty of the connecton could be, but also the greater the resource consumpton (e.g., bandwdth) and hence prce. Ths paper frst deals wth the Avalablty- Based Path Selecton (ABPS) problem, whch s to establsh a connecton over at most k > 0 (fully or partally) lnk-dsjont paths, for whch the avalablty s at least δ (0 < δ 1). Apart from consderng how to provde a relable connecton to customers, t s also mportant for network operators to determne the most vulnerable part of the network,.e., a subset of lnks wth hghest falure probablty whose removal wll dsconnect the network. The network operator could then replace/strengthen those lnks n order to ncrease the network relablty. Hence, ths paper also tackles ths so-called Network Vulnerablty Assessment (NVA) problem, whch s to fnd a set of network cuts for whch the falure probablty of the lnks n a cut belongng to that set s hghest. Our key contrbutons are as follows: We consder the Avalablty-Based Path Selecton (ABPS) problem n generc networks, Shared-Rsk Lnk Group (SRLG) networks and mult-layer networks. We prove that, n general, the ABPS problem cannot be approxmated n polynomal tme. We propose a polynomal-tme heurstc algorthm and an exact Integer Non-Lnear Program (INLP) to solve the ABPS problem. We compare, va smulatons, the proposed algorthms wth two exstng algorthms n terms of performance and runnng tme. We consder the Network Vulnerablty Assessment (NVA) problem n generc networks, SRLG networks and mult-layer networks. The remander of ths paper s organzed as follows. Related work s presented n Secton II. Secton III explans the calculaton of avalablty for dfferent path types: unprotected path, k fully lnk-dsjont and k partally lnk-dsjont. In Secton IV, we formally defne the Avalablty-Based Path Selecton (ABPS) problem n generc networks and analyze ts complexty. In Secton V and VI, we consder the ABPS problem n SRLG networks and mult-layer networks, respectvely. Secton VII presents our heurstc routng algorthm and an exact INLP. Secton VIII provdes our smulaton results. In Secton IX, we study the Network Vulnerablty Assessment problem n the three aforementoned types of networks. We conclude n Secton X. II. RELATED WORK Avalablty-aware routng under both statc and dynamc traffc demands has been extensvely nvestgated [1], [2], [3], [4], [5], [6]. When the traffc matrx s gven n advance (statc

2 traffc), Zhang et al. [3] present a mathematcal model to compute avalablty for dfferent protecton types (unprotected, dedcated protecton and shared protecton) for a gven statc traffc matrx. Furthermore, an Integer Lnear Program (ILP) and a heurstc algorthm are proposed to fnd avalabltyaware paths. Tornatore et al. [4] address the avalablty desgn problem: to accommodate a gven traffc matrx by usng shared/dedcated protecton paths. Song et al. [5] propose an avalablty-guaranteed routng algorthm, where dfferent protecton types are allowed. They defne a new cost functon for computng a backup path when the unprotected path fals to satsfy the avalablty requrement. She et al. [1] prove that for dedcated protecton, fndng two lnk-dsjont paths wth maxmal relablty (avalablty) s NP-hard. They also propose two heurstcs for that problem. Luo et al. [6] analyze the problem of protecton wth dfferent relablty, whch s to fnd one unprotected path or dedcated protecton path such that the cost of the whole path s mnmzed and the relablty requrement s satsfed. They subsequently propose an exact ILP as well as two approxmaton algorthms. However, the relablty (avalablty) calculaton n [6] s dfferent from the aforementoned papers, and assumes a sngle-lnk falure model. Assumng each lnk n the network has a falure probablty (=1-avalablty), Lee and Modano [2] mnmze the total falure probablty of unprotected, partally lnkdsjont and fully lnk-dsjont paths by establshng INLPs. They further transform the proposed INLPs to ILPs by usng lnear approxmatons. Dfferent from the aforementoned artcles, we target a more general problem, whch s to fnd at most k (fully or partally) lnk-dsjont paths for whch the avalablty requrement s satsfed. From the perspectve of network relablty calculaton, assumng each lnk n the network s assocated wth a falure probablty value (=1-avalablty), Provan and Ball [7] prove the problem of computng the probablty that the network stays connected s #P-complete 1. Karger [9] proposes a Fully Polynomal Randomzed Approxmaton Scheme (FPRAS) to solve ths problem. There s also work focusng on how to mathematcally model the avalablty of varous network topologes or dfferent protecton segments/paths. Zou et al. [10] nvestgate how to mathematcally calculate the avalablty of dfferent types of network topologes, e.g., tree topology, double star, crown or trple star. Tornatore et al. [11] mathematcally model the avalablty of segment protecton (SP). In segment protecton, a workng path (WP) can be parttoned nto several workng segments (WSs) and each WS s protected by a backup segment (BS). Moreover, they consder two SP cases, namely (1) overlap SP, where dfferent WSs can share the same lnk, and (2) no-overlap SP, where WSs are fully lnk-dsjont. By expressng the duallnk falure va a contnuous tme Markov chan, Mello et al. [12] approxmately estmate the (un)avalablty of the 1 Valant [8] shows that problems n ths class are at least as hard as NPcomplete problems. shared protecton path. Regardng SRLG networks, Hu [13] proves that the problem of fndng 2 SRLG-dsjont paths s NP-hard. To solve t, Hu [13] presents an exact ILP and Xu et al. [14] propose a trapavodance heurstc algorthm. However, the SRLG-dsjont routng problem s not the same as the one studed n ths paper, due to Eq. (11) n Secton V. Hence, the algorthms n [13] [14] cannot be used to effectvely solve our problem. In generc networks, the (s, t) Mn-Cut problem refers to parttonng the network nto two dsjont subsets such that nodes s and t are n dfferent subsets and the total weght of the cut lnks s mnmzed. Accordng to [15], ths problem can be solved by fndng the maxmum flow from s to t. There s also a lot of work on the Mn-Cut problem wth no specfed node pars (s, t). An enumeraton of classcal algorthms to solve the Mn-Cut problem can be found n [16]. III. CONNECTION AVAILABILITY The avalablty of a system s the fracton of tme the system s operatonal durng the entre servce tme. Lke [1], [2], [3], we frst assume that, n generc networks, the lnks avalabltes are uncorrelated/ndependent. If a connecton s carred by a sngle (unprotected) path, ts avalablty s equal to the path avalablty; f t s protected by k 2 dsjont paths, the avalablty wll be determned by these k protecton paths. The avalablty A j of a network component j can be calculated as [10]: A j = MT T F MT T F + MT T R where MT T F represents Mean Tme To Falure and MT T R denotes Mean Tme To Repar. We assume that the lnk avalablty s equal to the product of avalabltes of all ts components (e.g., amplfers). A. Lnk Falure Scenaros For smplcty, suppose there are two (fully) lnk-dsjont paths p 1 and p 2, and the avalablty of lnk l s denoted as A l (= 1 falure probablty), where 0 < A l 1, then ther total avalablty A 2 F D can be computed based on the followng scenaros: Sngle-lnk falure: Here t s assumed that all the lnks n the network have very low falure probablty. In ths context, a path p s avalablty (denoted by A p ) s equal to ts lowest traversed lnk avalablty (hghest falure probablty),.e., A p = mn l p A l. Usng two dsjont paths (whch s a conventonal survvablty mechansm) wll therefore lead to a total connecton avalablty of 1. However, ths approach only works when all the lnks are hghly relable. In Appendx A, we wll address the ABPS problem under the sngle-lnk falure scenaro. Multple lnk falures: Ths s a more general scenaro where at one certan pont n tme, several lnks n the network may fal smultaneously. Hence, for a path p, ts avalablty A p should take nto account all ts lnks avalabltes,.e., A p = l p A l. Consequently, A 2 F D = (1)

3 1 (1 A p1 )(1 A p2 ), whch ndcates the probablty that at least one path s avalable. In ths paper, we assume multple lnk falures may occur. B. End-to-End Path Avalablty If a path p contans the lnks l 1, l 2, l 3,..., l m, and ther correspondng avalabltes are denoted by A l1, A l2, A l3,..., A lm, then the avalablty of ths path (represented by A p ) s equal to 2 : A p = A l1 A l2 A l3 A lm (2) If we take the log of the lnk avalabltes, fndng the path wth the hghest avalablty turns nto a shortest path problem. When, for a sngle connecton, there are k 2 lnk-dsjont paths p 1, p 2,..., p k wth avalabltes represented by A p1, A p2,..., A pk, the connecton avalablty calculaton can be dvded nto two cases, namely: (1) fully lnk-dsjont paths: these k paths have no lnks n common, and (2) partally lnkdsjont paths: at least two of these k paths traverse at least one same lnk. In case (1), the avalablty (represented by A k F D ) s: A k F D = 1 + 0<<j<u k k (1 A p ) = =1 k A p =1 0<<j k A p A pj A pu + + ( 1) k 1 A u A u 1 a A v A w s b t A p A pj k =1 A p (3) Fg. 1: Network wth dfferent lnk avalablty values. If we use Eq. (3) to calculate the avalablty for the partally lnk dsjont case, the avalablty of the overlappng lnks wll be counted more than once. To amend ths, we use a new operator, whch s defned as follows: X 1 X 2 X k Y = { k =1 X f X = Y k =1 X Y otherwse where X 1, X 2,..., X k and Y represent dfferent lnk avalabltes. Therefore, the avalablty (represented by A k P D ) of k partally lnk-dsjont paths can be calculated as follows: 2 A network havng node and lnk avalabltes can be transformed to a drected network wth only lnk avalabltes, as done n [17]. Therefore, we assume the nodes have avalablty 1 n ths paper. (4) A k P D = 1 k (1 A p ) =1 = 1 (1 A p1 ) (1 A p2 ) (1 A pk ) (5) k = A p A p A pj + =1 0<<j<u k 0<<j k A p A pj A pu + + ( 1) k 1 k A p =1 where s used to denote the operatons of dfferent sets. Let us use an example to descrbe the dfference between case (1) and case (2), where k s set to 2 for smplcty. In Fg. 1 where the lnk avalablty s labeled on each lnk, paths s a t and s b t are fully lnk dsjont. Accordng to Eq. (3), ther avalablty s equal to: 1 (1 A u A w ) (1 A u A v ) =A u A w + A u A v A 2 u A w A v (6) On the other hand, paths s a t and s a b t are only partally lnk dsjont. Accordng to Eq. (5), the connecton avalablty can be calculated as follows: 1 (1 A u A w ) (1 A u A v ) =A u A w + A u A v A u A w A v (7) The followng theorem wll formalze the ntutve noton that f a set of paths p wth avalabltes A p have overlappng lnks that ther total avalablty s less than when those paths would have been fully lnk dsjont. Theorem 1: For gven A p, where 1 k, A k F D A k P D Ṗroof: A proof by mathematcal nducton: When k = 2, A 2 F D = A p 1 + A p2 A p1 A p2, and A 2 P D = A p1 +A p2 A p1 A p2. Snce A p1 A p2 A p1 A p2 accordng to Eq. (4) when 0 A p 1, the theorem s correct for k = 2. Assume when k = m the theorem s correct: m m A m F D = 1 (1 A p ) A m P D = 1 (1 A p ) (8) =1 =1 When k = m + 1, A m+1 F D = 1 (1 A p 1 ) (1 A p2 ) (1 A pm ) (1 A pm+1 ) and A m+1 P D = 1 (1 A p 1 ) (1 A p2 ) (1 A pm ) (1 A pm+1 ). Accordng to Eq. (8), we have: (1 A p1 ) (1 A p2 ) (1 A pm ) (1 A pm+1 ) (1 A p1 ) (1 A p2 ) (1 A pm ) (1 A pm+1 ) (9) Snce (1 A pm ) (1 A pm+1 ) (1 A pm ) (1 A pm+1 ), we have: (1 A p1 ) (1 A p2 ) (1 A pm ) (1 A pm+1 ) (1 A p1 ) (1 A p2 ) (1 A pm ) (1 A pm+1 ) (10) By mergng Eq. (9) and Eq. (10), we ascertan that A m+1 F D A m+1 P D.

4 IV. ABPS PROBLEM AND COMPLEXITY A. Problem Defnton Defnton 1: Gven a network represented by G(N, L) where N represents the set of N nodes and L denotes the set of L lnks, and each lnk l has ts own avalablty value A l. For a request represented by r(s, t, δ), where s and t denote the source and destnaton, respectvely, and δ (0 < δ 1) represents the avalablty requrement, establsh a connecton over at most k (partally) lnk-dsjont paths for whch the avalablty s at least δ. A varant, called the Avalablty-Based Backup Path Selecton (ABBPS) problem, s defned as: Defnton 2: Gven an exstng prmary path p from s to t and a requested avalablty δ, fnd at most k 1 paths that are fully or partally lnk-dsjont wth p, such that the avalablty of these k paths s no less than δ. B. Complexty Analyss In ths secton, we study the complexty of the ABPS problem n generc networks. For the case k = 1, by takng the log of the lnk avalabltes, the ABPS problem turns nto a shortest path problem, whch s polynomally solvable. Theorem 2: The ABPS problem s NP-hard for k 2. Proof: The case for partally lnk-dsjont paths can be reduced to the case of fully lnk-dsjont paths by a transformaton such as n Fg. 2. More specfcally, f we assume that all lnks n Fg. 2, except for (s, s ) and (t, t), have avalablty less than δ, then no lnk, except for (s, s ) and (t, t), can be an unprotected lnk n the soluton of the ABPS problem for the partally lnk dsjont case from s to t. Hence, solvng the fully lnk-dsjont ABPS problem from s to t s equvalent to solvng the partally lnk-dsjont ABPS problem from s to t. We therefore proceed to prove that the fully lnk dsjont varant for k = 2 s NP-hard. The proof for k > 2 follows analogously from the proof for k = 2. x a u 1 u1 v 1 v1 u 2 u 2 v 2 v 2 Fg. 3: A lobe for each x. 2 3 n u q u q m m Fg. 4: Lobes for all clauses. C two nodes y and z are created and a lnk connects z and y +1 wth avalablty of 1, where 0 < < m. We assume that s = x 1 and t = x n+1. Moreover, we draw a lnk (s, y 1 ) wth avalablty a and a lnk (z m, t) wth avalablty 1, where 0 < b < a 2 < 1. Fg. 4 depcts ths process. To relate the clause and varables n the constructed graph, we add the followng lnks: () lnks (y j, u k ) and (v k, z j) are added f the k-th occurrence of varable x exsts n clause C j ; or () lnks (y j, u k ) and (v k, z j) are added f the k-th occurrence of varable x exsts wth a negaton n the clause C j. For nstance, a network correspondng to 3SAT nstance (x 1 x 2 x 4 ) (x 1 x 2 x 3 ) (x 2 x 3 x 4 ) (x 1 x 3 x 4 ) s shown n Fg. 5. Based on the constructed graph, whch v q v q x +1 a Fg. 2: Reducton of ABPS problem from partally lnk dsjont to fully lnk dsjont. We frst ntroduce the NP-hard 3SAT problem [18] and then reduce the ABPS problem to t. The 3SAT problem s defned as: There s a boolean formula C 1 C 2...C m, where C denotes the -th clause. Each clause contans 3 varables wth an OR relaton. The queston s whether there s a truth assgnment to the varables that smultaneously satsfes all m clauses. Gven a 3SAT nstance, the graph constructon follows smlarly to [1]. Assume there are n varables n the 3SAT nstance. Frst, we create a lobe for each varable x, whch s shown n Fg. 3, where q represents the number of occurrences of varable x n all the clauses. The avalablty value for each lnk s also shown n Fg. 3, where 0 < b < 1. For each clause Fg. 5: Constructed graph that corresponds to (x 1 x 2 x 4 ) (x 1 x 2 x 3 ) (x 2 x 3 x 4 ) (x 1 x 3 x 4 ). corresponds to a gven 3SAT nstance, we are asked to solve the ABPS problem for k = 2 and δ = a + b q ab q, where q s the sum of occurrences for each varable n all the clauses,.e., q = n =1 q. Because one shortest path can at most have avalablty a, whch s less than δ, we have to fnd 2 lnkdsjont paths. Next, we wll prove that the fully lnk dsjont varant of the ABPS problem s NP-hard. 3SAT to ABPS: If there exsts a truth assgnment that satsfes all the clauses, then each clause j has (at least) one varable wth true or (negated) false assgnment to make ths clause true. Therefore, an upper subpath y j u k v k z j y j+1 or a lower subpath y j u k v k z j y j+1 wll be selected.

5 By concatenatng these m subpaths wth s y 1 and z m t we obtan one path (denoted by p 1 ) wth avalablty a. Snce each varable only has one truth assgnment, p 1 cannot traverse both the upper subpath and lower subpath n the same lobe. Subsequently, we can get another fully lnk-dsjont path p 2 : For each lobe (correspondng to varable x ), p 2 traverses the upper (lower) subpath wth avalablty of b q f p 1 goes through the lnk of lower (upper) subpath. The avalablty of p 2 s b q = b n =1 q, therefore p 1 and p 2 together have avalablty of a + b q ab q, whch satsfes the requrement δ. ABPS to 3SAT: If there are two fully lnk-dsjont paths from s to t wth avalablty no less than a + b q ab q, then one path must have avalablty a. To understand ths, assume that none of the two paths has avalablty a; wthout loss of generalty, we denote one path has avalablty of a c b e, where c can be ether 0 or 1 ndcatng whether lnk (s, y 1 ) has been traversed, and e > 0 s the number of lnks that have avalablty b. Snce there exsts only one lnk wth avalablty a, the other lnk-dsjont path has avalablty a c b f, where c s ether 0 or 1 meanng whether lnk (s, y 1 ) has been traversed and c + c 1, and f > 0 s the number of lnks whch have avalablty b. Hence, the avalablty of these two paths s a c b e + a c b f a c+c b e+f < b + b < a < δ, when b < a 2. Based on ths analyss, there must exst one path p 1 from s to t wth avalablty a, whch goes through (s, y 1 ) and (z m, t) and the other lnks wth avalablty of 1. To satsfy the avalablty requrement, there must also exst another fully lnk-dsjont path p 2 from s to t wth avalablty of no less than b q. For each lobe, p 2 should traverse ether the upper subpath or the lower subpath, otherwse p 1 and p 2 cannot be fully lnk dsjont. Therefore, p 2 wll traverse the (entre) lower subpath f p 1 goes through lnk (u k, v k ) n the upper subpath, and traverse the (entre) upper subpath f p 1 goes through lnk (u k, v k ) n the lower subpath for each lobe x. That s to say, p 1 cannot smultaneously traverse one lnk n the upper subpath and another lnk n the lower subpath for each same lobe. Consequently, p 1 ether goes va an upper subpath y j u k v k z j y j+1 to set varable x to true or va a lower subpath y j u k v k z j y j+1 to set varable x to false for clause j, where = 1, 2,..., n and j = 1, 2,..., m. Hence, all the m clauses can be smultaneously satsfed. Theorem 3: The ABBPS problem s NP-hard for k 2. Proof: For k 3, the ABBPS problem s equvalent to the ABPS problem for k 1 fully or partally lnk-dsjont paths, and hence NP-hard. In Appendx B, we prove that the ABBPS problem s also NP-hard for k = 2. We proceed to study the approxmablty of the ABPS problem. Theorem 4: The ABPS problem for k 2 cannot be approxmated to arbtrary degree n polynomal tme, unless P=NP. Proof: We can check n polynomal tme whether a sngle path can accommodate the requested avalablty. Hence, the theorem s equvalent to: for a request r(s, t, δ) and any constant number d > 1, there s no polynomal-tme algorthm that can fnd at least 2, but at most k, fully or partally lnkdsjont paths from s to t wth avalablty at least δ d. We prove the theorem for the fully lnk dsjont varant 3 of the ABPS problem for k = 2. We wll use a proof by contradcton and assume a polynomal-tme approxmaton algorthm A exsts for any d > 1. In the constructed graph based on the gven 3SAT nstance n Fg. 5 (also usng the same notaton and condtons), assume δ = a + b q ab q, so algorthm A can fnd two a+b fully lnk-dsjont paths wth avalablty at least q ab q d. Next, we prove that when 0 < b < a 2d, except for an exact soluton, there exsts no soluton wth avalablty no less than a+b q ab q d. If the exact soluton s not acheved by algorthm A, accordng to our prevous analyss, then one path must have avalablty of a c b e and the other path has avalablty of a c b f. Therefore, the avalablty of these two paths s equal to a c b e + a c b f a c+c b e+f. For a gven d, we have a c b e +a c b f a c+c b e+f < b+b = 2b < a d, when 0 < b < a 2d and 0 < a < 1. Therefore, under 0 < b <, except for an exact soluton, any two fully lnk-dsjont paths cannot have avalablty less than a+bq ab q d. To fulfll the assumpton, algorthm A has to fnd two lnk-dsjont paths wth avalablty a+b q ab q. In ths context, the fully lnk dsjont varant of the ABPS problem for k = 2 can be solved exactly n polynomal tme, whch s a contradcton. V. SHARED-RISK LINK GROUPS In ths secton, we assume two types of falures/avalabltes, namely Shared-Rsk Lnk Group (SRLG) falures and sngle lnk falures/avalabltes. A Shared-Rsk Lnk Group (SRLG) [16] reflects that a certan set/group of lnks n a network wll fal smultaneously. For nstance, n optcal networks, several fbers may resde n the same duct and a cut of the duct would cut all fbers n t. One duct n ths context corresponds to one dstnct SRLG. If each lnk s a sngle member of an SRLG, then no SRLGs exst. Hence the ABPS problem n SRLG networks ncludes as a specal case the ABPS problem n generc networks as dscussed n the prevous secton. Each lnk can belong to one or more SRLGs, and the lnks n the same SRLG wll smultaneously fal when the correspondng SRLG fals. The probablty of ths happenng (or not) s the SRLG falure (avalablty) probablty. We assume there are g SRLGs n the network G(N, L), and that the falure probablty of the -th SRLG (represented by srlg ) s denoted by π, for 1 g. For a partcular lnk l L, we denote by SR l the set of all SRLGs to whch l belongs. Dfferent from [2], where all SRLG events are assumed to be mutually exclusve, we assume that multple SRLG events may occur smultaneously. The avalablty of a sngle path should ncorporate the SRLG avalabltes as well as the lnk avalabltes. Consequently, the avalablty of path p can be calculated as: ;srlg p (1 π ) l p 3 The partally lnk dsjont varant follows analogously. a 2d A l (11)

6 where ;srlg p (1 π ) n Eq. (11) s the contrbuton of all the traversed SRLGs, whle l p A l s the avalablty of path p under the condton that all ts traversed SRLGs do not fal. For example, n Fg. 6, suppose there are three SRLGs n the network wth falure probabltes 0.1, 0.4 and 0.2, respectvely, and all the lnks have avalablty 0.9. We calculate the avalablty of path s a b t, whch traverses 2 SRLGs (srlg 1 and srlg 3 ): The probablty that both srlg 1 and srlg 3 do not fal s (1 0.1) (1 0.2). Under ths condton, all the lnks on path s a b t have avalablty 0.9 and therefore path s a b t has a total avalablty of (1 0.1) (1 0.2) (0.9) 3 = relable path s avalablty s = However, the optmal soluton s path s-b-t wth avalablty 0.45 n the physcal layer. The reason s that (s, a) and (a, t) n the logcal layer share the same lnk (s, b), whch leads to a lower avalablty value. Fg. 7(b) shows a smlar example wth Fg. 7(a), except that each lnk n the physcal layer has one addtonal wavelength number. In the absence of wavelength converson, t s requred that the lghtpath occupes the same wavelength on all lnks t traverses, whch s referred to as the wavelength-contnuty constrant n WDM-enabled networks. Now, suppose we want to fnd the most relable lghtpath from s to t, that obeys the wavelength-contnuty constrant. Clearly, f we are only aware of the lnks n the logcal layer, the result s path s-a-t wth avalablty = However, snce ths path s mapped to the path s-a-b-t n the physcal layer, t volates the wavelength-contnuty constrant. The optmal soluton s path s-a-t n the physcal layer va wavelength λ 1. Its avalablty s = Fg. 6: Avalablty calculaton n an SRLG network. Next, we wll prove that the sngle path varant of the ABPS problem n SRLG networks s NP-hard. To that end, we frst ntroduce the Mnmum Color Sngle-Path (MCSP) problem, whch s NP-hard [19]. Gven a network G(N, L), and gven the set of colors C = {c 1, c 2,..., c g } where g s the total number of colors n G, and gven the color {c l } of every lnk l L, the Mnmum Color Sngle-Path (MCSP) problem s to fnd a path from source node s to destnaton node t that uses the least amount of colors. Theorem 5: The ABPS problem s NP-hard n SRLG networks even for k = 1. Proof: Assume we have a network where all the lnks have avalablty 1 when ther SRLGs do not fal, and that there are g SRLGs wth the same falure probablty 1 g. Hence, a path s avalablty s only determned by the number of SRLGs t traverses. If we denote one SRLG by one partcular color, then the sngle-path ABPS problem n SRLG networks can be reduced to the MSCP problem. VI. MULTI-LAYER NETWORKS In mult-layer (e.g., IP-over-WDM) networks or overlay networks, the abstract lnks n the logcal layer are mapped to dfferent physcal lnks n the physcal layer. In ths context, two or more abstract lnks that contan the same physcal lnks may have correlated avalabltes or falure probabltes. Moreover, usually only the lnks n the logcal layer are known n mult-layer networks. Let us frst consder the example of mult-layer networks shown n Fg. 7. In Fg. 7(a), the avalablty s labeled on each lnk n the physcal layer, and the lnks n the logcal layer are mapped to the lnks n the physcal layer wth the greatest avalablty. Suppose we want to fnd a most relable unprotected path from s to t n Fg. 7(a). Snce we are only aware of the lnks n the logcal layer, we fnd that the most (a) Increasng jont avalablty (b) Decreasng jont avalablty Fg. 7: A mult-layer network example. For any two lnks l and m n the logcal layer of a multlayer network, we denote ther avalabltes by A l and A m, respectvely. Let us use as the actual jont avalablty value of these two lnks, and then we derve that A l A m can be greater than A l A m (Fg. 7(a)), or less than A l A m (Fg. 7(b)), or equal to A l A m. For the latter case, we say that l and m are uncorrelated, otherwse we say that l and m are correlated: f A l A m > A l A m, l and m are ncreasng correlated, else l and m are decreasng correlated. Analogously, the operator 1

7 can be used for more than two lnks. Next, we wll prove the NP-hardness of the ABPS problem for k = 1 n mult-layer networks. Theorem 6: The ABPS problem s NP-hard n mult-layer networks even for k = 1. Proof: When all the lnks are uncorrelated n multlayer networks, the ABPS wth k = 1 problem s solvable polynomal tme. SRLG networks can be regarded as a specal case of ncreasng correlaton n mult-layer networks, snce the lnks that share at least one common SRLG group (denote ths ( lnk set by L ) wll have a greater avalablty than (1 π ) A l ). Snce the sngle-path ABPS l L ;srlg l problem n SRLG networks s NP-hard, as we proved n Theorem 5, the sngle-path ABPS problem n mult-layer networks s also NP-hard. VII. HEURISTIC AND EXACT ALGORITHMS A. Heurstc Algorthm Algorthm 1 (G, s, t, δ, k, I) 1: Fnd one shortest path p 1, return t f the avalablty requrement s satsfed, otherwse go to Step 2. 2: ps 1, H p 1, P H, P b and Q 3: Whle ps k 4: P H 5: For each path ap P 6: P b P and counter 0 7: Whle counter I do 8: Randomly remove one lnk (u, v) ap and fnd one shortest path ψ u v from u to v. 9: If t succeeds then 10: Replace (u, v) wth ψ u v n ap, denote t as ap 11: P b. Remove(ap), P b. Add(ap ), ap ap 12: Fnd another lnk-dsjont path p 2 wth P b. 13: Return {p 2 } P b f δ s met. 14: For each lnk (u, v) ap 15: If ts avalablty s at least δ then 16: Q. Add((u, v)) 17: whle (Q ) do 18: (u, v) EXCT-MIN(Q) 19: Fnd a path p 3 whch shares (u, v) wth ap. 20: If (p 3 / P b ) and {p 3 } P b satsfy δ then 21: Return {p 3 } P b 22: else H Max Avalablty{H, {p 3 } P b } 23: counter counter : ps ps + 1 Our heurstc, called Mn-Mns Algorthm () to solve the ABPS problem n generc networks, SRLG networks and mult-layer networks, s presented n Algorthm 1. Snce we want to use as least (and no more than k) lnk-dsjont paths to satsfy the requested avalablty, we gradually ncrease the number of paths. The pseudo code to solve the ABPS problem n mult-layer networks s smlar to the one for generc networks, except for the path avalablty calculaton. Also, the pseudo code to solve the ABPS problem n SRLG networks s smlar to the one n generc networks, and we wll specfy the dfferences later. In what follows, we explan each step of the heurstc algorthm. We assgn lnk l L wth the weght of log(a l ) ( log( SR l(1 π ) A l ) for SRLG networks) n. If a shortest path (represented by p 1 ) n Step 1 fals to satsfy the avalablty requrement, we keep t as the ntal path flow. In Step 2, we use ps to record the number of already found lnk-dsjont paths. Intally ps s set to 1. H stores the already found ps lnk-dsjont paths, and t s ntally assgned p 1. Whle ps s no greater than k, Steps 3-24 contnue fndng a soluton. In Step 4, we assgn to P the already found paths H. Based on P, from Step 5 to Step 23, we each tme select one path ap from path set P. We also use a varable, denoted by counter n Algorthm 1, to record the number of teratons. Intally, counter s set to 0. As long as the number of teratons s less than an nput value I, Steps 7-23 proceed fndng a soluton based on ap and path set P b. The (sub)path from u to v found by the algorthm s denoted by ψ u v. In Step 8, we randomly remove one lnk (u, v) from ap, and we apply a shortest path algorthm from u to v to obtan a path ψ u v. By concatenatng subpath ψ u v and the lnks of path ap except for (u, v), we obtan a new path ap. Further, by substtutng ap wth ap n P b, we have a new path set P b. After that, the algorthm tres to fnd P b s fully lnkdsjont path n Step 12. When solvng the ABPS problem n SRLG networks, snce each SRLG only contrbutes once to the path avalablty calculaton, the lnk l s weght s set to log( {SR l \SR c } (1 π ) A l ) before runnng a shortest path algorthm n Step 12 (also the same for Step 19), where SR c are the common traversed SRLGs between lnk l and path set P b. If t fals to fnd p 2 or {p 2 } P b cannot satsfy the avalablty requrement, the algorthm tres to fnd a path whch s partally lnk dsjont wth ap (n Steps 13-22). The general dea s that we frst use a queue Q to store the lnks n ap whose avalablty s no less than δ n Steps After that, as long as Q s not empty n Steps 17-22, each tme the lnk wth the greatest avalablty n Q s extracted as the unprotected lnk (represented by (u, v)), and then we remove all the lnks traversed by ap except for (u, v). Subsequently, we fnd one shortest path ψ s u from s to u (f t exsts), and fnd another shortest path ψ v t from v to t (f t exsts). By concatenatng ψ s u, (u, v) and ψ v t, we can get a new path p 3, whch s partally lnk dsjont wth ap. If a and b denote dfferent sets of k > 1 lnk-dsjont paths, the functon Max Avalablty(a, b) n Step 22 returns the one wth greater avalablty. The tme complexty of can be computed as follows. Step 1 has a tme complexty of O(N log N + L). From Step 3 to Step 24, there are at most O(I) + O(2I) + + O(kI) = O(k 2 I) teratons before the algorthm termnates. Steps have a tme complexty of O(N) snce a path contans at most N 1 lnks and therefore Steps consume

8 O(N(N log N + L)) tme. Fnally, the whole tme complexty of s O(k 2 IN(N log N + L)). B. Exact INLP Formulaton In ths subsecton, we present an exact Integer Non-Lnear Program (INLP) to solve the ABPS problem n generc, SRLG and mult-layer networks. We frst solve the ABPS problem n generc networks and start by explanng the requred notaton and varables. INLP notaton: r(s, t, δ): Traffc request, wth source s, destnaton t and requested avalablty δ. A,j : Avalablty of lnk (, j). g : The total number of SRLGs. π,j m : The falure probablty of the m-th SRLG f lnk (, j) belongs to t, otherwse t s 0. INLP varable: P r,u,j : Boolean varable equal to 1 f lnk (, j) s traversed by path u (1 u k) for request r; 0 otherwse. Flow conservaton constrants: (,j) L N P r,u,j (j,) L 1 u k Avalablty constrant: k u=1 (,j) L P r,u j, = ( 1 P r,u,j + P r,j A,j ) ( mn 1 P r,u,j + P r,u,j 1, 1, 0, = s = t otherwse ) r,v A,j, 1 P,j + P r,v,j A,j (12) 1 u<v k (,j) L + + ( 1) k 1 r,u mn (1 P,j + P r,u,j A,j) δ (13) 1 u k (,j) L When both the flow conservaton constrant (Eq. (12)) and the avalablty constrant (Eq. (13)) are satsfed, an optmal soluton s found by the INLP, otherwse there s no soluton. There s no objectve (needed) n the proposed INLP, but one could nclude the objectve of mnmzng the number of paths (or lnks) used. Eq. (12) accounts for the flow conservaton for each of the at most k paths. For a partcular u th path (1 u k), t ensures that () for the source node s of request r, the outgong traffc for each request s 1; () for the destnaton node t of request r, the ncomng traffc s 1; and () for an ntermedate node whch s nether source nor destnaton, ts ncomng traffc s equal to the outgong traffc. Eq. (13) ensures that ether the found sngle unprotected path or the (partally) lnk-dsjont paths should have avalablty no less than δ, accordng to the avalablty calculaton of k lnkdsjont paths n Eqs. (3) and (5). Snce the overlapped lnk s avalablty n the partally lnk-dsjont calculaton accordng to Eq. (5) can only be counted once, we take the mnmum value of the varables P r,u,j for each lnk and then take the product over all the lnks for (partally) lnk-dsjont paths. We also note that Eq. (13) can smultaneously calculate the avalablty of the fully lnk dsjont varant, partally lnk dsjont varant and the unprotected varant. For nstance when k = 2, Eq. (13) becomes: or (1 P r,1,j + P r,1,j A,j) + (1 P r,2,j + P r,2,j A,j) (,j) L (,j) L mn(1 P r,1,j + P r,1,j A,j, 1 P r,2,j + P r,2,j A,j) δ (14) (,j) L When P r,1,j = P r,2,j for all (, j) L, Eq. (14) s equal to (1 P r,1,j + P r,1,j A,j) δ (,j) L (,j) L (1 P r,2,j + P r,2,j A,j) δ whch s the avalablty constrant for a sngle unprotected path. To solve the ABPS problem n SRLG networks, we need to slghtly modfy Eq. (13) (and keep flow conversaton constrants Eq. (12) the same) by usng 1 m g k mn u=1 ( ) r,u 1 mn (1 P,j + P r,u,j (,j) L πm,j ) to multply the left sde of Eq. (13), whch s the non-falure probablty of the SRLGs whch at most k lnk-dsjont paths have traversed. In mult-layer networks, the avalablty of a subset of lnks may not be equal to the product of ther avalabltes. Therefore, we need one more functon f( L ), whch can return the jont avalablty of a subset of lnks L n mult-layer networks. The parameter of ths functon s a 0/1 lnk vector whch contans L elements, where 1 denotes the lnk s present to be calculated and 0 means t s not. Consequently, to solve the ABPS problem n multlayer networks exactly, we could replace the operator wth f() n Eq. (13), and keep all the other notaton and constrants the same as for the generc networks case. A. Smulaton Setup VIII. SIMULATION RESULTS Fg. 8: USA carrer backbone network. We conduct smulatons on two networks, one s USANet, dsplayed n Fg. 8, whch s a realstc carrer backbone network consstng of 24 nodes and 43 lnks, and the other s GÉANT, shown n Fg. 9, whch s the pan-european communcatons nfrastructure servng Europe s research and

9 Fg. 9: GÉANT pan-european research network. educaton communty consstng of 40 nodes and 63 lnks. The smulaton deals wth the ABPS problem n generc, SRLG and mult-layer networks. For generc networks, we assume the avalablty of fber lnks s dstrbuted among the set {0.99, 0.999, }, wth a proporton of 1:1:2. Based on the same lnk avalabltes, n SRLG networks we assume that there are n total 5 SRLG events wth the falure probabltes 0.001, 0.002, 0.003, and 0.005, respectvely. Each lnk has randomly been assgned to at most 3 SRLG events. Based on the same (ndvdual) lnk avalabltes as for generc 1 networks, n mult-layer networks, 3 of lnks are ncreasng correlated, and t follows that A l A m A n = max(a l, A m,, A n ); 1 3 of lnks are decreasng correlated, and t follows that A l A m A n = (A l A m A n ) 2 ; and the other 1 3 of lnks are uncorrelated. For all these three networks, snce we want to compare the ablty of fndng paths for the algorthms, the capacty s set to nfnty. We vary the number of traffc requests from 100 to The source and destnaton of each request are randomly selected, and each request has nfnte holdng tme. The requested avalablty ncludes two cases: () general avalablty requrement case: the avalablty s randomly dstrbuted among the set {0.98, 0.99, 0.995, 0.997, 0.999}; () hgh avalablty requrement case: the avalablty s randomly dstrbuted among the set {0.9995, , , , }, by whch we want to challenge the algorthm to fnd feasble paths under more dffcult condtons. Consderng the practcal tme complexty and the exstng proposed algorthms that only focus on fndng two lnk-dsjont paths, we choose k = 2. We set I n to be logn n these two networks (5 n USANet and 6 n GÉANT, respectvely). Under the same weght allocaton wth our algorthm, we compare the proposed heurstc and exact INLP wth two heurstcs: Two-step Relablty Algorthm () and Maxmal-Relablty Algorthm (), whch are proposed n [1]. frst calculates a shortest path, and then calculates (f t exsts) another shortest path after removng the lnks traversed by the frst path. apples Suurballe s algorthm [20] to calculate a par of two lnkdsjont paths that have mnmum weght. Both algorthms frst apply a shortest path algorthm to check whether an unprotected path soluton exts. The smulaton s run on a desktop PC wth 3.00 GHz CPU and 4 GB memory. We use IBM ILOG CPLEX 12.6 to mplement the proposed INLP and C# to mplement the heurstc algorthms. B. Results We frst evaluate the performance of the algorthms n terms of Acceptance Rato (AR) n generc networks. Acceptance rato (AR) s defned as the percentage of the number of accepted requests over all the requests. We frst analyze the general avalablty requrement case: In USANet, all the algorthms acheved an AR of 1. We therefore omt the fgure of the general avalablty performance for USANet. However, ths s not the case for the GÉANT topology. From Fg. 10(a), we can see that the performance of all algorthms s under Snce GÉANT s not as well connected as USANet s, some nodes n GÉANT only have degree one (e.g., nodes 3, 8, etc.), f a one-degree node becomes the source or the destnaton of a certan request, the request can only be served by partal protecton (or a sngle unprotected path). In ths context, a feasble path may not exst n GÉANT, whch wll result n blockng. In terms of performance, the INLP acheves the hghest AR. On the other hand, shows a hgher AR than the other two heurstcs and (Fg. 10(a)). For the hgh avalablty requrement scenaro (shown n Fgs. 10(b) and 10(c)), as expected, the AR of all these algorthms s lower than n the general avalablty requrement case. In ths scenaro, the INLP requres more tme to fnd a soluton, especally when a soluton does not exst. In order to let the INLP return the result n a reasonable tme, we set the tme lmt for t to serve one request to 50 mnutes. Due to ths reason, we can see that INLP has the lowest AR n USANet and often second hghest AR n GÉANT. Meanwhle, stll has the hghest AR n most of the cases. The tme lmt for the INLP s even more constranng n the case of SRLG networks, leadng to a very poor performance for SRLG networks. We have therefore omtted the results of the INLP n SRLG networks. Snce the optmal soluton rarely exsts n the hgh avalablty requrement case, we only provde the smulaton results for the heurstc algorthms n the general avalablty requrement case. Moreover, to have a far comparson, we compare our algorthms wth and a modfed [2], whch s a heurstc routng algorthm proposed for probablstc SRLG networks. Its man dea s that after fndng the frst shortest path, the remanng lnk weghts should be adjusted (We slghtly change ts lnk weght adjustment to be the same wth the Step 12 of for a farer comparson), and then to fnd another lnk-dsjont shortest path. Fg. 11 shows that the proposed heurstc algorthm stll acheves hgher AR than these two algorthms. Smlar to SRLG networks, the exact INLP n mult-layer networks s very tme consumng. We therefore omt the results of the INLP n mult-layer networks. Fg. 12 provdes the

10 Acceptance rato (AR) INLP traffc demand Acceptance rato (AR) INLP* traffc demand Acceptance rato (AR) INLP* traffc demand (a) GÉANT (b) USANet (c) GÉANT Fg. 10: AR of four algorthms n generc networks: (a) general avalablty requrement. (USANet has been omtted snce all 4 algorthms always acheved an AR of 1.); (b)-(c) hgh avalablty requrement, * max 50 mns per request. Acceptance rato (AR) traffc demand (a) USANet Acceptance rato (AR) traffc demand (b) GÉANT Fg. 11: AR of the heurstc algorthms n SRLG networks for general avalablty requrement. results for all three heurstcs n the two networks for both general and hgh avalablty requrement scenaros. It can be seen that acheves the hghest AR compared to the other two heurstcs. TABLE I: Runnng tmes per request for four algorthms (ms). Networks INLP USA Generc (General δ) GÉANT Generc (General δ) USA Generc (Hgh δ) GÉANT Generc (Hgh δ) USA SRLG (General δ) > GÉANT SRLG (General δ) > USA Mult-layer (General δ) > GÉANT Mult-layer (General δ) > USA Mult-layer (Hgh δ) > GÉANT Mult-layer (Hgh δ) > Fnally, n Table I, we present the (average) runnng tmes per request for these four algorthms n generc, SRLG and mult-layer networks. It shows that the INLP s sgnfcantly more tme consumng than the three polynomal-tme heurstcs. On the other hand, has only a slghtly hgher runnng tme than and, but t pays off by havng a hgher AR as shown n Fgs Another observaton s that, for the same algorthm n the same network, the runnng tme s hgher for the hgh avalablty requrement case than n the general avalablty requrement case. IX. NETWORK VULNERABILITY ASSESSMENT A. Problem Defnton and Complexty Analyss Fndng the most vulnerable part of a network as well as the prevously consdered problem of avalablty-based path selecton are both mportant elements for network robustness. In ths secton, we study the (s,t) Network Vulnerablty Assessment (NVA) problem. That s, fnd one or a set of (equal-weght) network cuts whose falure probablty of the lnks n the cut s hghest. A network cut refers to a set of lnks, whose removal wll result n the dsconnecton of the network. Formally, the NVA problem can be defned as follows: Defnton 3: The (s,t) Network Vulnerablty Assessment (NVA) problem: Gven s a network G(N, L), and each lnk l L s assocated wth a falure probablty f l = 1 A l. Gven a source s and a target t, fnd an s t cut C for whch l C f l s maxmzed. In case there are multple cuts of hghest weght all of them should be returned. When the node par (s,t) s not specfed, we denote ths problem as the NVA problem, whch can be solved by solvng the (s,t) NVA problem at most N 1 tmes. Therefore, these two problems share the same hardness. We use log(f l ) for the weght of lnk l n the network. In generc networks, the NVA problem can be solved by fndng all the mn-cuts n O(L 2 N + N 2 L) tme accordng to [21]. On the other hand, we wll prove that the NVA problem n SRLG networks s NP-hard. Recall that n SRLG networks, ntroduced n Secton V, two types of falures/avalabltes should be ncorporated, namely Shared-Rsk Lnk Group (SRLG) falures and sngle lnk falures/avalabltes, therefore the probablty that all the lnks belongng to path p fal smultaneously (denoted by F (p)) can be calculated as: 1 (1 π ) + f l 1 (1 π ) l p ;srlg p ;srlg p l p f l (15)

11 Acceptance rato (AR) traffc demand (a) USANet Acceptance rato (AR) traffc demand (b) GÉANT Acceptance rato (AR) traffc demand (c) USANet Acceptance rato (AR) traffc demand (d) GÉANT Fg. 12: AR of heurstcs n mult-layer networks: (a)-(b) general avalablty requrement; (c)-(d) hgh avalablty requrement. Please note that Eq. (15) s dfferent from the falure probablty of a path n SRLG networks: 1 Eq. (11). Before we prove the NP-hardness of the NVA problem, let us frst study the Maxmum Color Path Selecton (MCPS) problem. Contrary to the MCSP problem, the MCPS problem s to fnd a sngle path such that t uses the largest amount of colors. Theorem 7: The Maxmum Color Path Selecton (MCPS) problem s NP-hard. Proof: We dstngush two cases, namely, (1) all the lnks have exclusve colors,.e., there does not exst any color that s shared/overlapped by two or more lnks, and (2) two or more lnks may contan the same color(s). Case (1): The MCPS problem s equvalent to the NP-hard Longest Path problem [18], whch s to fnd a path from s to t such that ts weght s maxmzed. Fg. 14: Example network. x 1 x 2 x 3 x n s y 1 y 2 y 3 y n-1 t z 1 z 2 z 3 z n Fg. 13: Reducton of the MCPS problem to the Dsjont Connectng Paths problem. Case (2): We frst ntroduce the Dsjont Connectng Paths problem [18]. Gven a network G(N, L), and a collecton of dsjont node pars (s 1, t 1 ), (s 2, t 2 ),..., (s z, t z ), does G contan z mutually lnk-dsjont paths, one connectng s and t for each, 1 z? Ths problem s proved NP-hard when z 3, and we reduce the MCPS problem to t. In Fg. 13, assume that there are n total g colors and lnks (a, b) and (c, d) share 0 < x < g common colors, but they together contan g dstnct colors. Except for these two lnks, the other lnks are assgned 0 < y < g colors, but there does not exst two or more lnks contanng g dstnct colors. In ths context, fndng a path from s to t wth the largest amount of colors s equvalent to fndng three mutually lnk-dsjont paths between three node pars (s, a), (b, c) and (d, t). Theorem 8: The Network Vulnerablty Assessment (NVA) problem n SRLG Networks s NP-hard. Fg. 15: Reducton of the (s,t) NVA problem n SRLG networks to the MCPS problem. Proof: It s equvalent to prove that the (s,t) NVA problem s NP-hard. In Fg. 14, assume all lnk avalabltes are 1 and all lnks have non-zero SRLG falure probabltes, except for lnks (x, x +1 ) and (z, z +1 ) whch have 0 SRLGs, where 1 n 1. Assume there are g SRLG events n total, and each SRLG event occurs wth a probablty of 1 g. In ths context, for a path p wthout lnks (x, x +1 ) and (z, z +1 ), accordng to Eq. (15), F (p) can be calculated as: 1 (1 1 g )m (16) where m s the total number of dstnct SRLGs traversed by p. Therefore, to maxmze Eq. (16) one needs to maxmze m,.e., to fnd a path wth the greatest probablty that all ts traversed lnks fal smultaneously. Ths s equvalent to fndng a path havng the largest number of dstnct SRLGs. We want to solve the (x 1,z 1 ) NVA problem. Based on Fg. 14, we frst

12 derve Fg. 15 wth the same nodes except that we add one more node s. We regard s = y 0, and t = y n. The lnk weght n Fg. 15 s set as follows: (y 1, x ) and (y 1, z ) have 0 SRLGs, whle (x, z ) and (z, y ) have the same SRLGs as n Fg. 14, where 1 n. In Fg. 15, we are asked to solve the MCPS problem from the source s to the destnaton t. Snce we want to fnd a cut that separates x 1 and z 1, any cuts n the form of (x, y ) and (y, z ), where 1 n are not part of the optmal soluton. Moreover, consderng the lnk n the form of (x j, x j+1 ) or (y j, y j+1 ) has 0 SRLGs and sngle lnk avalablty 1, whch means ts avalablty s 1 or falure probablty s 0, t cannot lead to the optmal soluton as well. Based on the above observatons, any feasble cut C should contan one lnk ether (x, y ) or (y, z ), for all 1 n. We prove n the followng that the (s,t) NVA problem n Fg. 14 can be reduced to the MCPS problem n Fg. 15 n polynomal tme, where, for smplcty, we assume only one optmal cut exsts n the (s,t) NVA problem. (s,t) NVA to MCPS: An optmal soluton of the (s,t) NVA problem should be composed of ether (x, y ) or (y, z ), where 1 n. Let C NV A reflect the set of lnks n the optmal soluton. Because C NV A has the largest amount of dstnct SRLGs, C NV A together wth (s, x 1 ) or (s, z 1 ) forms a path from s to t wth the maxmum number of SRLGs. MCPS to (s,t) NVA: Let R MCP S denote the set of lnks n the optmal soluton of the MCPS problem. Because R MCP S has the largest amount of SRLGs, let C NV A = R MCP S \{(y, x +1 ), (y, z +1 )}. Consderng that the lnks (y, x +1 ) and (y, z +1 ) have 0 SRLGs, C NV A also has the largest amount of SRLGs. Therefore solvng the MCPS problem yelds a soluton to the (s,t) NVA problem. Theorem 9: The NVA problem n mult-layer networks s NP-hard. Proof: For any two lnks l and m n a mult-layer network, we have that A l A m = (1 f l ) (1 f m ) = 1 f m f l + f l f m. Smlar to Theorem 6, we can also prove that t s NP-hard to fnd one path from the source to the destnaton n mult-layer networks such that the probablty that all ts lnks fal s maxmzed. Subsequently, we can prove that the NVA problem n mult-layer networks s NP-hard, whch follows analogously wth Theorem 8. We therefore omt the detals. X. CONCLUSION The avalablty of a connecton represents how relably a connecton can carry data from a source to a destnaton. In ths paper, we have frst studed the Avalablty-Based Path Selecton (ABPS) problem, whch s to establsh a connecton over at most k (fully or partally) lnk-dsjont paths, such that the total avalablty s no less than δ (0 < δ 1). We have proved that, n general, the ABPS problem s NP-hard and cannot be approxmated n polynomal tme for k 2, unless P=NP. We have further proved that n SRLG networks and mult-layer networks, even the sngle-path (k = 1) varant of the ABPS problem s NP-hard. We have proposed a polynomal-tme heurstc algorthm and an exact INLP to solve the ABPS problem n generc networks, SRLG networks and mult-layer networks. Va smulatons, we have found that our heurstc algorthm outperforms two exstng algorthms n terms of acceptance rato wth only slghtly hgher runnng tme. On the other hand, the runnng tme of the exact INLP s sgnfcantly larger (by several orders of magntude) than the runnng tme of the heurstc algorthms. Fnally, we have proved that the Network Vulnerablty Assessment (NVA) problem, whch s to fnd a cut of the network for whch the falure probablty of all ts lnks s hghest, s solvable n polynomal tme n generc networks, but NP-hard to solve n SRLG networks and mult-layer networks. REFERENCES [1] Q. She, X. Huang, and J. Jue, How relable can two-path protecton be? IEEE/ACM Transactons on Networkng, vol. 18, no. 3, pp , [2] H.-W. Lee, E. Modano, and K. Lee, Dverse routng n networks wth probablstc falures, IEEE/ACM Transactons on Networkng, vol. 18, no. 6, pp , [3] J. Zhang, K. Zhu, H. Zang, N. Matloff, and B. Mukherjee, Avalabltyaware provsonng strateges for dfferentated protecton servces n wavelength-convertble wdm mesh networks, IEEE/ACM Transactons on Networkng, vol. 15, no. 5, pp , [4] M. Tornatore, G. Maer, and A. Pattavna, Avalablty desgn of optcal transport networks, IEEE Journal on Selected Areas n Communcatons, vol. 23, no. 8, pp , [5] L. Song, J. Zhang, and B. Mukherjee, Dynamc provsonng wth avalablty guarantee for dfferentated servces n survvable mesh networks, IEEE Journal on Selected Areas n Communcatons, vol. 25, no. 3, pp , [6] H. Luo, L. L, and H. Yu, Routng connectons wth dfferentated relablty requrements n wdm mesh networks, IEEE/ACM Transactons on Networkng, vol. 17, no. 1, pp , [7] J. S. Provan and M. O. Ball, The complexty of countng cuts and of computng the probablty that a graph s connected, SIAM Journal on Computng, vol. 12, no. 4, pp , [8] L. G. Valant, The complexty of enumeraton and relablty problems, SIAM Journal on Computng, vol. 8, no. 3, pp , [9] D. R. Karger, A randomzed fully polynomal tme approxmaton scheme for the all-termnal network relablty problem, SIAM revew, vol. 43, no. 3, pp , [10] W. Zou, M. Janć, R. Kooj, and F. Kupers, On the avalablty of networks, n Proc. of BroadBand Europe, Belgum, December, [11] M. Tornatore, M. Carcagn, and A. Pattavna, Avalablty formulatons for segment protecton, IEEE Transactons on Communcatons, vol. 58, no. 4, pp , [12] D. A. A. Mello, D. Schupke, and H. Waldman, A matrx-based analytcal approach to connecton unavalablty estmaton n shared backup path protecton, IEEE Communcatons Letters, vol. 9, no. 9, pp , [13] J. Q. Hu, Dverse routng n optcal mesh networks, IEEE Transactons on Communcatons, vol. 51, no. 3, pp , [14] D. Xu, Y. Xong, C. Qao, and G. L, Trap avodance and protecton schemes n networks wth shared rsk lnk groups, Journal of Lghtwave Technology, vol. 21, no. 11, pp , [15] L. R. Ford and D. R. Fulkerson, Maxmal flow through a network, Canadan Journal of Mathematcs, vol. 8, no. 3, pp , [16] F. A. Kupers, An overvew of algorthms for network survvablty, ISRN Communcatons and Networkng, vol. 2012, p. 19, [17] G. Dantzg and D. R. Fulkerson, On the max flow mn cut theorem of networks, Lnear nequaltes and related systems, vol. 38, pp , [18] M. R. Garey and D. S. Johnson, Computers and Intractablty: A Gude to the Theory of NP-Completeness. New York, NY, USA: W. H. Freeman & Co., 1979.

13 [19] S. Yuan, S. Varma, and J. Jue, Mnmum-color path problems for relablty n mesh networks, INFOCOM 2005, pp , [20] J. W. Suurballe and R. E. Tarjan, A quck method for fndng shortest pars of dsjont paths, Networks, vol. 14, no. 2, pp , [21] H. Nagamoch, K. Nshmura, and T. Ibarak, Computng all small cuts n an undrected network, SIAM Journal on Dscrete Mathematcs, vol. 10, no. 3, pp , APPENDIX A SINGLE-LINK FAILURE In ths secton, we assume that only 1 lnk n the network can fal at a tme. Apart from that, we also assume that two successve falures occur wth a tme dfference longer than the traversal tme of any path n the network. For ths scenaro, we present a polynomal-tme algorthm to solve the ABPS problem. More specfcally, n Algorthm 2, when k = 1, n Steps 1-2, we elmnate the lnks wth avalablty less than δ, such that we obtan a new graph G where each lnk has avalablty at least δ. Subsequently, by runnng Djkstra s shortest path algorthm on G from s to t, we can solve the ABPS problem for k = 1. When k 2, f the optmal soluton conssts of k fully lnk-dsjont paths, then 2 fully lnk-dsjont paths also exst and under the sngle-lnk falure scenaro have avalablty s 1, whch s optmal. Hence, by applyng Suurballe s algorthm [20] n Step 3, the soluton can be found, f t exsts. When the optmal soluton conssts of k partally lnkdsjont paths, then 2 partally lnk-dsjont paths are also enough. The reason s that the avalablty of partally lnkdsjont paths s decded by one unprotected lnk (say l). Hence, t suffces to fnd k = 2 partally lnk-dsjont paths. In Steps 4-5, for each lnk (u, v) whose avalablty s no less than δ, we create another (parallel) lnk between u and v wth the same avalablty. After that, we call Suurballe s algorthm [20] from s to t. Snce n the optmal soluton the unprotected lnk has avalablty at least δ, by creatng the parallel lnks whose avalablty s at least δ, the paths p 1 and p 2 returned by Suurballe s algorthm [20] are two fully lnk-dsjont paths. After that, f p 1 and p 2 traverse the parallel lnks, we then merge these two lnks nto one lnk. Ths knd of lnk reflects the unprotected lnk n the optmal soluton. On the other hand, the lnks whose avalablty s less than δ are protected n the returned soluton because of the correctness of Suurballe s algorthm. Therefore, an optmal soluton can be found by algorthm 2. APPENDIX B HARDNESS OF THE ABBPS PROBLEM FOR K=2 For some varants, lke the fully lnk dsjont case, the ABBPS problem s polynomally solvable 4, however t s NPhard n ts general settng. Theorem 10: The partally lnk-dsjont ABBPS problem for k = 2 s NP-hard. Proof: We provde a proof for when k = 2. As t s shown n Fg. 16, assume we are gven a path (denoted by GP) s-a-bc-d-t wth the avalablty labeled on each lnk and all the other Algorthm 2 ABPSSngleLnkFalure(G, s, t, δ, k) 1: Elmnate the lnks wth avalablty less than δ on G, thereby obtanng a new graph G. 2: Run Djkstra s algorthm on G from s to t. If the soluton s found then return the result; Else f k > 1, contnue; Otherwse output there s no soluton. 3: Run Suurballe s algorthm [20] on G from s to t. Return the result f the soluton s found, otherwse contnue. 4: Create another (parallel) lnk between u and v wth the same avalablty A (u,v), for each (u, v) L f A (u,v) δ. The graph s denoted as G. 5: Run Suurballe s algorthm [20] on G from s to t. Return the result f the soluton s found, otherwse output there s no soluton. Fg. 16: Reducton of the ABBPS problem wth partally lnk dsjont paths to the Dsjont Connectng Paths problem. lnks have an avalablty of 1. We now want to fnd a partally lnk-dsjont path wth GP such that ther combned avalablty s no less than 1. Snce the requested avalablty s 1, only lnk (a, b) and lnk (c, d) can be unprotected n an optmal soluton. Suppose that when lnk (a, b) s elmnated, there do not exst paths from node s to nodes b, c, d and t, and when lnk (c, d) s elmnated, there are no paths from node b to nodes d and t. In ths context, to solve the partally lnk-dsjont ABBPS problem for k = 2, both (a, b) and (c, d) should be unprotected n the optmal soluton. Ths s equvalent to fndng three pars of lnk-dsjont paths between node pars (s, a), (b, c) and (d, t) (.e., the Dsjont Connectng Paths problem). Hence, solvng the partally lnk-dsjont ABBPS problem for k = 2 yelds a soluton to the NP-hard Dsjont Connectng Paths problem. 4 Trvally, by lookng for an unprotected path wth maxmum avalablty n the network where all the lnks from the prmary path are excluded.

Luby s Alg. for Maximal Independent Sets using Pairwise Independence

Luby s Alg. for Maximal Independent Sets using Pairwise Independence Lecture Notes for Randomzed Algorthms Luby s Alg. for Maxmal Independent Sets usng Parwse Independence Last Updated by Erc Vgoda on February, 006 8. Maxmal Independent Sets For a graph G = (V, E), an ndependent

More information

The Greedy Method. Introduction. 0/1 Knapsack Problem

The Greedy Method. Introduction. 0/1 Knapsack Problem The Greedy Method Introducton We have completed data structures. We now are gong to look at algorthm desgn methods. Often we are lookng at optmzaton problems whose performance s exponental. For an optmzaton

More information

Recurrence. 1 Definitions and main statements

Recurrence. 1 Definitions and main statements Recurrence 1 Defntons and man statements Let X n, n = 0, 1, 2,... be a MC wth the state space S = (1, 2,...), transton probabltes p j = P {X n+1 = j X n = }, and the transton matrx P = (p j ),j S def.

More information

PAS: A Packet Accounting System to Limit the Effects of DoS & DDoS. Debish Fesehaye & Klara Naherstedt University of Illinois-Urbana Champaign

PAS: A Packet Accounting System to Limit the Effects of DoS & DDoS. Debish Fesehaye & Klara Naherstedt University of Illinois-Urbana Champaign PAS: A Packet Accountng System to Lmt the Effects of DoS & DDoS Debsh Fesehaye & Klara Naherstedt Unversty of Illnos-Urbana Champagn DoS and DDoS DDoS attacks are ncreasng threats to our dgtal world. Exstng

More information

J. Parallel Distrib. Comput.

J. Parallel Distrib. Comput. J. Parallel Dstrb. Comput. 71 (2011) 62 76 Contents lsts avalable at ScenceDrect J. Parallel Dstrb. Comput. journal homepage: www.elsever.com/locate/jpdc Optmzng server placement n dstrbuted systems n

More information

8.5 UNITARY AND HERMITIAN MATRICES. The conjugate transpose of a complex matrix A, denoted by A*, is given by

8.5 UNITARY AND HERMITIAN MATRICES. The conjugate transpose of a complex matrix A, denoted by A*, is given by 6 CHAPTER 8 COMPLEX VECTOR SPACES 5. Fnd the kernel of the lnear transformaton gven n Exercse 5. In Exercses 55 and 56, fnd the mage of v, for the ndcated composton, where and are gven by the followng

More information

A New Paradigm for Load Balancing in Wireless Mesh Networks

A New Paradigm for Load Balancing in Wireless Mesh Networks A New Paradgm for Load Balancng n Wreless Mesh Networks Abstract: Obtanng maxmum throughput across a network or a mesh through optmal load balancng s known to be an NP-hard problem. Desgnng effcent load

More information

) of the Cell class is created containing information about events associated with the cell. Events are added to the Cell instance

) of the Cell class is created containing information about events associated with the cell. Events are added to the Cell instance Calbraton Method Instances of the Cell class (one nstance for each FMS cell) contan ADC raw data and methods assocated wth each partcular FMS cell. The calbraton method ncludes event selecton (Class Cell

More information

What is Candidate Sampling

What is Candidate Sampling What s Canddate Samplng Say we have a multclass or mult label problem where each tranng example ( x, T ) conssts of a context x a small (mult)set of target classes T out of a large unverse L of possble

More information

benefit is 2, paid if the policyholder dies within the year, and probability of death within the year is ).

benefit is 2, paid if the policyholder dies within the year, and probability of death within the year is ). REVIEW OF RISK MANAGEMENT CONCEPTS LOSS DISTRIBUTIONS AND INSURANCE Loss and nsurance: When someone s subject to the rsk of ncurrng a fnancal loss, the loss s generally modeled usng a random varable or

More information

Virtual Network Embedding with Coordinated Node and Link Mapping

Virtual Network Embedding with Coordinated Node and Link Mapping Vrtual Network Embeddng wth Coordnated Node and Lnk Mappng N. M. Mosharaf Kabr Chowdhury Cherton School of Computer Scence Unversty of Waterloo Waterloo, Canada Emal: nmmkchow@uwaterloo.ca Muntasr Rahan

More information

A Design Method of High-availability and Low-optical-loss Optical Aggregation Network Architecture

A Design Method of High-availability and Low-optical-loss Optical Aggregation Network Architecture A Desgn Method of Hgh-avalablty and Low-optcal-loss Optcal Aggregaton Network Archtecture Takehro Sato, Kuntaka Ashzawa, Kazumasa Tokuhash, Dasuke Ish, Satoru Okamoto and Naoak Yamanaka Dept. of Informaton

More information

Energy Efficient Routing in Ad Hoc Disaster Recovery Networks

Energy Efficient Routing in Ad Hoc Disaster Recovery Networks Energy Effcent Routng n Ad Hoc Dsaster Recovery Networks Gl Zussman and Adran Segall Department of Electrcal Engneerng Technon Israel Insttute of Technology Hafa 32000, Israel {glz@tx, segall@ee}.technon.ac.l

More information

1 Example 1: Axis-aligned rectangles

1 Example 1: Axis-aligned rectangles COS 511: Theoretcal Machne Learnng Lecturer: Rob Schapre Lecture # 6 Scrbe: Aaron Schld February 21, 2013 Last class, we dscussed an analogue for Occam s Razor for nfnte hypothess spaces that, n conjuncton

More information

Forecasting the Demand of Emergency Supplies: Based on the CBR Theory and BP Neural Network

Forecasting the Demand of Emergency Supplies: Based on the CBR Theory and BP Neural Network 700 Proceedngs of the 8th Internatonal Conference on Innovaton & Management Forecastng the Demand of Emergency Supples: Based on the CBR Theory and BP Neural Network Fu Deqang, Lu Yun, L Changbng School

More information

Data Broadcast on a Multi-System Heterogeneous Overlayed Wireless Network *

Data Broadcast on a Multi-System Heterogeneous Overlayed Wireless Network * JOURNAL OF INFORMATION SCIENCE AND ENGINEERING 24, 819-840 (2008) Data Broadcast on a Mult-System Heterogeneous Overlayed Wreless Network * Department of Computer Scence Natonal Chao Tung Unversty Hsnchu,

More information

Minimal Coding Network With Combinatorial Structure For Instantaneous Recovery From Edge Failures

Minimal Coding Network With Combinatorial Structure For Instantaneous Recovery From Edge Failures Mnmal Codng Network Wth Combnatoral Structure For Instantaneous Recovery From Edge Falures Ashly Joseph 1, Mr.M.Sadsh Sendl 2, Dr.S.Karthk 3 1 Fnal Year ME CSE Student Department of Computer Scence Engneerng

More information

How To Plan A Network Wide Load Balancing Route For A Network Wde Network (Network)

How To Plan A Network Wide Load Balancing Route For A Network Wde Network (Network) Network-Wde Load Balancng Routng Wth Performance Guarantees Kartk Gopalan Tz-cker Chueh Yow-Jan Ln Florda State Unversty Stony Brook Unversty Telcorda Research kartk@cs.fsu.edu chueh@cs.sunysb.edu yjln@research.telcorda.com

More information

Module 2 LOSSLESS IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur

Module 2 LOSSLESS IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur Module LOSSLESS IMAGE COMPRESSION SYSTEMS Lesson 3 Lossless Compresson: Huffman Codng Instructonal Objectves At the end of ths lesson, the students should be able to:. Defne and measure source entropy..

More information

On the Optimal Control of a Cascade of Hydro-Electric Power Stations

On the Optimal Control of a Cascade of Hydro-Electric Power Stations On the Optmal Control of a Cascade of Hydro-Electrc Power Statons M.C.M. Guedes a, A.F. Rbero a, G.V. Smrnov b and S. Vlela c a Department of Mathematcs, School of Scences, Unversty of Porto, Portugal;

More information

Agile Traffic Merging for Data Center Networks. Qing Yi and Suresh Singh Portland State University, Oregon June 10 th, 2014

Agile Traffic Merging for Data Center Networks. Qing Yi and Suresh Singh Portland State University, Oregon June 10 th, 2014 Agle Traffc Mergng for Data Center Networks Qng Y and Suresh Sngh Portland State Unversty, Oregon June 10 th, 2014 Agenda Background and motvaton Power optmzaton model Smulated greedy algorthm Traffc mergng

More information

Period and Deadline Selection for Schedulability in Real-Time Systems

Period and Deadline Selection for Schedulability in Real-Time Systems Perod and Deadlne Selecton for Schedulablty n Real-Tme Systems Thdapat Chantem, Xaofeng Wang, M.D. Lemmon, and X. Sharon Hu Department of Computer Scence and Engneerng, Department of Electrcal Engneerng

More information

General Auction Mechanism for Search Advertising

General Auction Mechanism for Search Advertising General Aucton Mechansm for Search Advertsng Gagan Aggarwal S. Muthukrshnan Dávd Pál Martn Pál Keywords game theory, onlne auctons, stable matchngs ABSTRACT Internet search advertsng s often sold by an

More information

Extending Probabilistic Dynamic Epistemic Logic

Extending Probabilistic Dynamic Epistemic Logic Extendng Probablstc Dynamc Epstemc Logc Joshua Sack May 29, 2008 Probablty Space Defnton A probablty space s a tuple (S, A, µ), where 1 S s a set called the sample space. 2 A P(S) s a σ-algebra: a set

More information

v a 1 b 1 i, a 2 b 2 i,..., a n b n i.

v a 1 b 1 i, a 2 b 2 i,..., a n b n i. SECTION 8.4 COMPLEX VECTOR SPACES AND INNER PRODUCTS 455 8.4 COMPLEX VECTOR SPACES AND INNER PRODUCTS All the vector spaces we have studed thus far n the text are real vector spaces snce the scalars are

More information

Feature selection for intrusion detection. Slobodan Petrović NISlab, Gjøvik University College

Feature selection for intrusion detection. Slobodan Petrović NISlab, Gjøvik University College Feature selecton for ntruson detecton Slobodan Petrovć NISlab, Gjøvk Unversty College Contents The feature selecton problem Intruson detecton Traffc features relevant for IDS The CFS measure The mrmr measure

More information

Fault tolerance in cloud technologies presented as a service

Fault tolerance in cloud technologies presented as a service Internatonal Scentfc Conference Computer Scence 2015 Pavel Dzhunev, PhD student Fault tolerance n cloud technologes presented as a servce INTRODUCTION Improvements n technques for vrtualzaton and performance

More information

On Multi-controller Placement Optimization in Software Defined Networking - based WANs

On Multi-controller Placement Optimization in Software Defined Networking - based WANs On Mult-controller Placement Optmzaton n Software Defned Networkng - based WANs Eugen Borcoc, Radu Badea, Serban Georgca Obreja, Marus Vochn Unversty POLITEHNICA of Bucharest - UPB Bucharest, Romana eugen.borcoc@elcom.pub.ro,

More information

An Alternative Way to Measure Private Equity Performance

An Alternative Way to Measure Private Equity Performance An Alternatve Way to Measure Prvate Equty Performance Peter Todd Parlux Investment Technology LLC Summary Internal Rate of Return (IRR) s probably the most common way to measure the performance of prvate

More information

Heuristic Static Load-Balancing Algorithm Applied to CESM

Heuristic Static Load-Balancing Algorithm Applied to CESM Heurstc Statc Load-Balancng Algorthm Appled to CESM 1 Yur Alexeev, 1 Sher Mckelson, 1 Sven Leyffer, 1 Robert Jacob, 2 Anthony Crag 1 Argonne Natonal Laboratory, 9700 S. Cass Avenue, Argonne, IL 60439,

More information

Traffic State Estimation in the Traffic Management Center of Berlin

Traffic State Estimation in the Traffic Management Center of Berlin Traffc State Estmaton n the Traffc Management Center of Berln Authors: Peter Vortsch, PTV AG, Stumpfstrasse, D-763 Karlsruhe, Germany phone ++49/72/965/35, emal peter.vortsch@ptv.de Peter Möhl, PTV AG,

More information

An Interest-Oriented Network Evolution Mechanism for Online Communities

An Interest-Oriented Network Evolution Mechanism for Online Communities An Interest-Orented Network Evoluton Mechansm for Onlne Communtes Cahong Sun and Xaopng Yang School of Informaton, Renmn Unversty of Chna, Bejng 100872, P.R. Chna {chsun,yang}@ruc.edu.cn Abstract. Onlne

More information

THE DISTRIBUTION OF LOAN PORTFOLIO VALUE * Oldrich Alfons Vasicek

THE DISTRIBUTION OF LOAN PORTFOLIO VALUE * Oldrich Alfons Vasicek HE DISRIBUION OF LOAN PORFOLIO VALUE * Oldrch Alfons Vascek he amount of captal necessary to support a portfolo of debt securtes depends on the probablty dstrbuton of the portfolo loss. Consder a portfolo

More information

How To Understand The Results Of The German Meris Cloud And Water Vapour Product

How To Understand The Results Of The German Meris Cloud And Water Vapour Product Ttel: Project: Doc. No.: MERIS level 3 cloud and water vapour products MAPP MAPP-ATBD-ClWVL3 Issue: 1 Revson: 0 Date: 9.12.1998 Functon Name Organsaton Sgnature Date Author: Bennartz FUB Preusker FUB Schüller

More information

+ + + - - This circuit than can be reduced to a planar circuit

+ + + - - This circuit than can be reduced to a planar circuit MeshCurrent Method The meshcurrent s analog of the nodeoltage method. We sole for a new set of arables, mesh currents, that automatcally satsfy KCLs. As such, meshcurrent method reduces crcut soluton to

More information

Institute of Informatics, Faculty of Business and Management, Brno University of Technology,Czech Republic

Institute of Informatics, Faculty of Business and Management, Brno University of Technology,Czech Republic Lagrange Multplers as Quanttatve Indcators n Economcs Ivan Mezník Insttute of Informatcs, Faculty of Busness and Management, Brno Unversty of TechnologCzech Republc Abstract The quanttatve role of Lagrange

More information

Open Access A Load Balancing Strategy with Bandwidth Constraint in Cloud Computing. Jing Deng 1,*, Ping Guo 2, Qi Li 3, Haizhu Chen 1

Open Access A Load Balancing Strategy with Bandwidth Constraint in Cloud Computing. Jing Deng 1,*, Ping Guo 2, Qi Li 3, Haizhu Chen 1 Send Orders for Reprnts to reprnts@benthamscence.ae The Open Cybernetcs & Systemcs Journal, 2014, 8, 115-121 115 Open Access A Load Balancng Strategy wth Bandwdth Constrant n Cloud Computng Jng Deng 1,*,

More information

A Probabilistic Theory of Coherence

A Probabilistic Theory of Coherence A Probablstc Theory of Coherence BRANDEN FITELSON. The Coherence Measure C Let E be a set of n propostons E,..., E n. We seek a probablstc measure C(E) of the degree of coherence of E. Intutvely, we want

More information

Logical Development Of Vogel s Approximation Method (LD-VAM): An Approach To Find Basic Feasible Solution Of Transportation Problem

Logical Development Of Vogel s Approximation Method (LD-VAM): An Approach To Find Basic Feasible Solution Of Transportation Problem INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME, ISSUE, FEBRUARY ISSN 77-866 Logcal Development Of Vogel s Approxmaton Method (LD- An Approach To Fnd Basc Feasble Soluton Of Transportaton

More information

Energy Conserving Routing in Wireless Ad-hoc Networks

Energy Conserving Routing in Wireless Ad-hoc Networks Energy Conservng Routng n Wreless Ad-hoc Networks Jae-Hwan Chang and Leandros Tassulas Department of Electrcal and Computer Engneerng & Insttute for Systems Research Unversty of Maryland at College ark

More information

How To Solve A Problem In A Powerline (Powerline) With A Powerbook (Powerbook)

How To Solve A Problem In A Powerline (Powerline) With A Powerbook (Powerbook) MIT 8.996: Topc n TCS: Internet Research Problems Sprng 2002 Lecture 7 March 20, 2002 Lecturer: Bran Dean Global Load Balancng Scrbe: John Kogel, Ben Leong In today s lecture, we dscuss global load balancng

More information

On the Interaction between Load Balancing and Speed Scaling

On the Interaction between Load Balancing and Speed Scaling On the Interacton between Load Balancng and Speed Scalng Ljun Chen, Na L and Steven H. Low Engneerng & Appled Scence Dvson, Calforna Insttute of Technology, USA Abstract Speed scalng has been wdely adopted

More information

Joint Scheduling of Processing and Shuffle Phases in MapReduce Systems

Joint Scheduling of Processing and Shuffle Phases in MapReduce Systems Jont Schedulng of Processng and Shuffle Phases n MapReduce Systems Fangfe Chen, Mural Kodalam, T. V. Lakshman Department of Computer Scence and Engneerng, The Penn State Unversty Bell Laboratores, Alcatel-Lucent

More information

An MILP model for planning of batch plants operating in a campaign-mode

An MILP model for planning of batch plants operating in a campaign-mode An MILP model for plannng of batch plants operatng n a campagn-mode Yanna Fumero Insttuto de Desarrollo y Dseño CONICET UTN yfumero@santafe-concet.gov.ar Gabrela Corsano Insttuto de Desarrollo y Dseño

More information

Number of Levels Cumulative Annual operating Income per year construction costs costs ($) ($) ($) 1 600,000 35,000 100,000 2 2,200,000 60,000 350,000

Number of Levels Cumulative Annual operating Income per year construction costs costs ($) ($) ($) 1 600,000 35,000 100,000 2 2,200,000 60,000 350,000 Problem Set 5 Solutons 1 MIT s consderng buldng a new car park near Kendall Square. o unversty funds are avalable (overhead rates are under pressure and the new faclty would have to pay for tself from

More information

On File Delay Minimization for Content Uploading to Media Cloud via Collaborative Wireless Network

On File Delay Minimization for Content Uploading to Media Cloud via Collaborative Wireless Network On Fle Delay Mnmzaton for Content Uploadng to Meda Cloud va Collaboratve Wreless Network Ge Zhang and Yonggang Wen School of Computer Engneerng Nanyang Technologcal Unversty Sngapore Emal: {zh0001ge, ygwen}@ntu.edu.sg

More information

Enabling P2P One-view Multi-party Video Conferencing

Enabling P2P One-view Multi-party Video Conferencing Enablng P2P One-vew Mult-party Vdeo Conferencng Yongxang Zhao, Yong Lu, Changja Chen, and JanYn Zhang Abstract Mult-Party Vdeo Conferencng (MPVC) facltates realtme group nteracton between users. Whle P2P

More information

Project Networks With Mixed-Time Constraints

Project Networks With Mixed-Time Constraints Project Networs Wth Mxed-Tme Constrants L Caccetta and B Wattananon Western Australan Centre of Excellence n Industral Optmsaton (WACEIO) Curtn Unversty of Technology GPO Box U1987 Perth Western Australa

More information

Multi-Resource Fair Allocation in Heterogeneous Cloud Computing Systems

Multi-Resource Fair Allocation in Heterogeneous Cloud Computing Systems 1 Mult-Resource Far Allocaton n Heterogeneous Cloud Computng Systems We Wang, Student Member, IEEE, Ben Lang, Senor Member, IEEE, Baochun L, Senor Member, IEEE Abstract We study the mult-resource allocaton

More information

DISTRIBUTED storage systems have been becoming increasingly

DISTRIBUTED storage systems have been becoming increasingly 268 IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 28, NO. 2, FEBRUARY 2010 Cooperatve Recovery of Dstrbuted Storage Systems from Multple Losses wth Network Codng Yuchong Hu, Ynlong Xu, Xaozhao

More information

How To Solve An Onlne Control Polcy On A Vrtualzed Data Center

How To Solve An Onlne Control Polcy On A Vrtualzed Data Center Dynamc Resource Allocaton and Power Management n Vrtualzed Data Centers Rahul Urgaonkar, Ulas C. Kozat, Ken Igarash, Mchael J. Neely urgaonka@usc.edu, {kozat, garash}@docomolabs-usa.com, mjneely@usc.edu

More information

Dynamic Fleet Management for Cybercars

Dynamic Fleet Management for Cybercars Proceedngs of the IEEE ITSC 2006 2006 IEEE Intellgent Transportaton Systems Conference Toronto, Canada, September 17-20, 2006 TC7.5 Dynamc Fleet Management for Cybercars Fenghu. Wang, Mng. Yang, Ruqng.

More information

A Replication-Based and Fault Tolerant Allocation Algorithm for Cloud Computing

A Replication-Based and Fault Tolerant Allocation Algorithm for Cloud Computing A Replcaton-Based and Fault Tolerant Allocaton Algorthm for Cloud Computng Tork Altameem Dept of Computer Scence, RCC, Kng Saud Unversty, PO Box: 28095 11437 Ryadh-Saud Araba Abstract The very large nfrastructure

More information

8 Algorithm for Binary Searching in Trees

8 Algorithm for Binary Searching in Trees 8 Algorthm for Bnary Searchng n Trees In ths secton we present our algorthm for bnary searchng n trees. A crucal observaton employed by the algorthm s that ths problem can be effcently solved when the

More information

Effective Network Defense Strategies against Malicious Attacks with Various Defense Mechanisms under Quality of Service Constraints

Effective Network Defense Strategies against Malicious Attacks with Various Defense Mechanisms under Quality of Service Constraints Effectve Network Defense Strateges aganst Malcous Attacks wth Varous Defense Mechansms under Qualty of Servce Constrants Frank Yeong-Sung Ln Department of Informaton Natonal Tawan Unversty Tape, Tawan,

More information

PSYCHOLOGICAL RESEARCH (PYC 304-C) Lecture 12

PSYCHOLOGICAL RESEARCH (PYC 304-C) Lecture 12 14 The Ch-squared dstrbuton PSYCHOLOGICAL RESEARCH (PYC 304-C) Lecture 1 If a normal varable X, havng mean µ and varance σ, s standardsed, the new varable Z has a mean 0 and varance 1. When ths standardsed

More information

ANALYZING THE RELATIONSHIPS BETWEEN QUALITY, TIME, AND COST IN PROJECT MANAGEMENT DECISION MAKING

ANALYZING THE RELATIONSHIPS BETWEEN QUALITY, TIME, AND COST IN PROJECT MANAGEMENT DECISION MAKING ANALYZING THE RELATIONSHIPS BETWEEN QUALITY, TIME, AND COST IN PROJECT MANAGEMENT DECISION MAKING Matthew J. Lberatore, Department of Management and Operatons, Vllanova Unversty, Vllanova, PA 19085, 610-519-4390,

More information

Schedulability Bound of Weighted Round Robin Schedulers for Hard Real-Time Systems

Schedulability Bound of Weighted Round Robin Schedulers for Hard Real-Time Systems Schedulablty Bound of Weghted Round Robn Schedulers for Hard Real-Tme Systems Janja Wu, Jyh-Charn Lu, and We Zhao Department of Computer Scence, Texas A&M Unversty {janjaw, lu, zhao}@cs.tamu.edu Abstract

More information

Answer: A). There is a flatter IS curve in the high MPC economy. Original LM LM after increase in M. IS curve for low MPC economy

Answer: A). There is a flatter IS curve in the high MPC economy. Original LM LM after increase in M. IS curve for low MPC economy 4.02 Quz Solutons Fall 2004 Multple-Choce Questons (30/00 ponts) Please, crcle the correct answer for each of the followng 0 multple-choce questons. For each queston, only one of the answers s correct.

More information

Network Aware Load-Balancing via Parallel VM Migration for Data Centers

Network Aware Load-Balancing via Parallel VM Migration for Data Centers Network Aware Load-Balancng va Parallel VM Mgraton for Data Centers Kun-Tng Chen 2, Chen Chen 12, Po-Hsang Wang 2 1 Informaton Technology Servce Center, 2 Department of Computer Scence Natonal Chao Tung

More information

Forecasting the Direction and Strength of Stock Market Movement

Forecasting the Direction and Strength of Stock Market Movement Forecastng the Drecton and Strength of Stock Market Movement Jngwe Chen Mng Chen Nan Ye cjngwe@stanford.edu mchen5@stanford.edu nanye@stanford.edu Abstract - Stock market s one of the most complcated systems

More information

CHOLESTEROL REFERENCE METHOD LABORATORY NETWORK. Sample Stability Protocol

CHOLESTEROL REFERENCE METHOD LABORATORY NETWORK. Sample Stability Protocol CHOLESTEROL REFERENCE METHOD LABORATORY NETWORK Sample Stablty Protocol Background The Cholesterol Reference Method Laboratory Network (CRMLN) developed certfcaton protocols for total cholesterol, HDL

More information

Formulating & Solving Integer Problems Chapter 11 289

Formulating & Solving Integer Problems Chapter 11 289 Formulatng & Solvng Integer Problems Chapter 11 289 The Optonal Stop TSP If we drop the requrement that every stop must be vsted, we then get the optonal stop TSP. Ths mght correspond to a ob sequencng

More information

INVESTIGATION OF VEHICULAR USERS FAIRNESS IN CDMA-HDR NETWORKS

INVESTIGATION OF VEHICULAR USERS FAIRNESS IN CDMA-HDR NETWORKS 21 22 September 2007, BULGARIA 119 Proceedngs of the Internatonal Conference on Informaton Technologes (InfoTech-2007) 21 st 22 nd September 2007, Bulgara vol. 2 INVESTIGATION OF VEHICULAR USERS FAIRNESS

More information

How Sets of Coherent Probabilities May Serve as Models for Degrees of Incoherence

How Sets of Coherent Probabilities May Serve as Models for Degrees of Incoherence 1 st Internatonal Symposum on Imprecse Probabltes and Ther Applcatons, Ghent, Belgum, 29 June 2 July 1999 How Sets of Coherent Probabltes May Serve as Models for Degrees of Incoherence Mar J. Schervsh

More information

The Development of Web Log Mining Based on Improve-K-Means Clustering Analysis

The Development of Web Log Mining Based on Improve-K-Means Clustering Analysis The Development of Web Log Mnng Based on Improve-K-Means Clusterng Analyss TngZhong Wang * College of Informaton Technology, Luoyang Normal Unversty, Luoyang, 471022, Chna wangtngzhong2@sna.cn Abstract.

More information

2008/8. An integrated model for warehouse and inventory planning. Géraldine Strack and Yves Pochet

2008/8. An integrated model for warehouse and inventory planning. Géraldine Strack and Yves Pochet 2008/8 An ntegrated model for warehouse and nventory plannng Géraldne Strack and Yves Pochet CORE Voe du Roman Pays 34 B-1348 Louvan-la-Neuve, Belgum. Tel (32 10) 47 43 04 Fax (32 10) 47 43 01 E-mal: corestat-lbrary@uclouvan.be

More information

Sngle Snk Buy at Bulk Problem and the Access Network

Sngle Snk Buy at Bulk Problem and the Access Network A Constant Factor Approxmaton for the Sngle Snk Edge Installaton Problem Sudpto Guha Adam Meyerson Kamesh Munagala Abstract We present the frst constant approxmaton to the sngle snk buy-at-bulk network

More information

Can Auto Liability Insurance Purchases Signal Risk Attitude?

Can Auto Liability Insurance Purchases Signal Risk Attitude? Internatonal Journal of Busness and Economcs, 2011, Vol. 10, No. 2, 159-164 Can Auto Lablty Insurance Purchases Sgnal Rsk Atttude? Chu-Shu L Department of Internatonal Busness, Asa Unversty, Tawan Sheng-Chang

More information

A Lyapunov Optimization Approach to Repeated Stochastic Games

A Lyapunov Optimization Approach to Repeated Stochastic Games PROC. ALLERTON CONFERENCE ON COMMUNICATION, CONTROL, AND COMPUTING, OCT. 2013 1 A Lyapunov Optmzaton Approach to Repeated Stochastc Games Mchael J. Neely Unversty of Southern Calforna http://www-bcf.usc.edu/

More information

Network Security Situation Evaluation Method for Distributed Denial of Service

Network Security Situation Evaluation Method for Distributed Denial of Service Network Securty Stuaton Evaluaton Method for Dstrbuted Denal of Servce Jn Q,2, Cu YMn,2, Huang MnHuan,2, Kuang XaoHu,2, TangHong,2 ) Scence and Technology on Informaton System Securty Laboratory, Bejng,

More information

On the Interaction between Load Balancing and Speed Scaling

On the Interaction between Load Balancing and Speed Scaling On the Interacton between Load Balancng and Speed Scalng Ljun Chen and Na L Abstract Speed scalng has been wdely adopted n computer and communcaton systems, n partcular, to reduce energy consumpton. An

More information

Allocating Collaborative Profit in Less-than-Truckload Carrier Alliance

Allocating Collaborative Profit in Less-than-Truckload Carrier Alliance J. Servce Scence & Management, 2010, 3: 143-149 do:10.4236/jssm.2010.31018 Publshed Onlne March 2010 (http://www.scrp.org/journal/jssm) 143 Allocatng Collaboratve Proft n Less-than-Truckload Carrer Allance

More information

Optimization of network mesh topologies and link capacities for congestion relief

Optimization of network mesh topologies and link capacities for congestion relief Optmzaton of networ mesh topologes and ln capactes for congeston relef D. de Vllers * J.M. Hattngh School of Computer-, Statstcal- and Mathematcal Scences Potchefstroom Unversty for CHE * E-mal: rwddv@pu.ac.za

More information

Lecture 3: Force of Interest, Real Interest Rate, Annuity

Lecture 3: Force of Interest, Real Interest Rate, Annuity Lecture 3: Force of Interest, Real Interest Rate, Annuty Goals: Study contnuous compoundng and force of nterest Dscuss real nterest rate Learn annuty-mmedate, and ts present value Study annuty-due, and

More information

Support Vector Machines

Support Vector Machines Support Vector Machnes Max Wellng Department of Computer Scence Unversty of Toronto 10 Kng s College Road Toronto, M5S 3G5 Canada wellng@cs.toronto.edu Abstract Ths s a note to explan support vector machnes.

More information

Calculation of Sampling Weights

Calculation of Sampling Weights Perre Foy Statstcs Canada 4 Calculaton of Samplng Weghts 4.1 OVERVIEW The basc sample desgn used n TIMSS Populatons 1 and 2 was a two-stage stratfed cluster desgn. 1 The frst stage conssted of a sample

More information

BERNSTEIN POLYNOMIALS

BERNSTEIN POLYNOMIALS On-Lne Geometrc Modelng Notes BERNSTEIN POLYNOMIALS Kenneth I. Joy Vsualzaton and Graphcs Research Group Department of Computer Scence Unversty of Calforna, Davs Overvew Polynomals are ncredbly useful

More information

1. Fundamentals of probability theory 2. Emergence of communication traffic 3. Stochastic & Markovian Processes (SP & MP)

1. Fundamentals of probability theory 2. Emergence of communication traffic 3. Stochastic & Markovian Processes (SP & MP) 6.3 / -- Communcaton Networks II (Görg) SS20 -- www.comnets.un-bremen.de Communcaton Networks II Contents. Fundamentals of probablty theory 2. Emergence of communcaton traffc 3. Stochastc & Markovan Processes

More information

行 政 院 國 家 科 學 委 員 會 補 助 專 題 研 究 計 畫 成 果 報 告 期 中 進 度 報 告

行 政 院 國 家 科 學 委 員 會 補 助 專 題 研 究 計 畫 成 果 報 告 期 中 進 度 報 告 行 政 院 國 家 科 學 委 員 會 補 助 專 題 研 究 計 畫 成 果 報 告 期 中 進 度 報 告 畫 類 別 : 個 別 型 計 畫 半 導 體 產 業 大 型 廠 房 之 設 施 規 劃 計 畫 編 號 :NSC 96-2628-E-009-026-MY3 執 行 期 間 : 2007 年 8 月 1 日 至 2010 年 7 月 31 日 計 畫 主 持 人 : 巫 木 誠 共 同

More information

Traffic-light a stress test for life insurance provisions

Traffic-light a stress test for life insurance provisions MEMORANDUM Date 006-09-7 Authors Bengt von Bahr, Göran Ronge Traffc-lght a stress test for lfe nsurance provsons Fnansnspetonen P.O. Box 6750 SE-113 85 Stocholm [Sveavägen 167] Tel +46 8 787 80 00 Fax

More information

A hybrid global optimization algorithm based on parallel chaos optimization and outlook algorithm

A hybrid global optimization algorithm based on parallel chaos optimization and outlook algorithm Avalable onlne www.ocpr.com Journal of Chemcal and Pharmaceutcal Research, 2014, 6(7):1884-1889 Research Artcle ISSN : 0975-7384 CODEN(USA) : JCPRC5 A hybrd global optmzaton algorthm based on parallel

More information

Analysis of Energy-Conserving Access Protocols for Wireless Identification Networks

Analysis of Energy-Conserving Access Protocols for Wireless Identification Networks From the Proceedngs of Internatonal Conference on Telecommuncaton Systems (ITC-97), March 2-23, 1997. 1 Analyss of Energy-Conservng Access Protocols for Wreless Identfcaton etworks Imrch Chlamtac a, Chara

More information

An Adaptive Cross-layer Bandwidth Scheduling Strategy for the Speed-Sensitive Strategy in Hierarchical Cellular Networks

An Adaptive Cross-layer Bandwidth Scheduling Strategy for the Speed-Sensitive Strategy in Hierarchical Cellular Networks An Adaptve Cross-layer Bandwdth Schedulng Strategy for the Speed-Senstve Strategy n erarchcal Cellular Networks Jong-Shn Chen #1, Me-Wen #2 Department of Informaton and Communcaton Engneerng ChaoYang Unversty

More information

Power-of-Two Policies for Single- Warehouse Multi-Retailer Inventory Systems with Order Frequency Discounts

Power-of-Two Policies for Single- Warehouse Multi-Retailer Inventory Systems with Order Frequency Discounts Power-of-wo Polces for Sngle- Warehouse Mult-Retaler Inventory Systems wth Order Frequency Dscounts José A. Ventura Pennsylvana State Unversty (USA) Yale. Herer echnon Israel Insttute of echnology (Israel)

More information

To Fill or not to Fill: The Gas Station Problem

To Fill or not to Fill: The Gas Station Problem To Fll or not to Fll: The Gas Staton Problem Samr Khuller Azarakhsh Malekan Julán Mestre Abstract In ths paper we study several routng problems that generalze shortest paths and the Travelng Salesman Problem.

More information

Chapter 4 ECONOMIC DISPATCH AND UNIT COMMITMENT

Chapter 4 ECONOMIC DISPATCH AND UNIT COMMITMENT Chapter 4 ECOOMIC DISATCH AD UIT COMMITMET ITRODUCTIO A power system has several power plants. Each power plant has several generatng unts. At any pont of tme, the total load n the system s met by the

More information

Dominant Resource Fairness in Cloud Computing Systems with Heterogeneous Servers

Dominant Resource Fairness in Cloud Computing Systems with Heterogeneous Servers 1 Domnant Resource Farness n Cloud Computng Systems wth Heterogeneous Servers We Wang, Baochun L, Ben Lang Department of Electrcal and Computer Engneerng Unversty of Toronto arxv:138.83v1 [cs.dc] 1 Aug

More information

POLYSA: A Polynomial Algorithm for Non-binary Constraint Satisfaction Problems with and

POLYSA: A Polynomial Algorithm for Non-binary Constraint Satisfaction Problems with and POLYSA: A Polynomal Algorthm for Non-bnary Constrant Satsfacton Problems wth and Mguel A. Saldo, Federco Barber Dpto. Sstemas Informátcos y Computacón Unversdad Poltécnca de Valenca, Camno de Vera s/n

More information

Omega 39 (2011) 313 322. Contents lists available at ScienceDirect. Omega. journal homepage: www.elsevier.com/locate/omega

Omega 39 (2011) 313 322. Contents lists available at ScienceDirect. Omega. journal homepage: www.elsevier.com/locate/omega Omega 39 (2011) 313 322 Contents lsts avalable at ScenceDrect Omega journal homepage: www.elsever.com/locate/omega Supply chan confguraton for dffuson of new products: An ntegrated optmzaton approach Mehd

More information

When Network Effect Meets Congestion Effect: Leveraging Social Services for Wireless Services

When Network Effect Meets Congestion Effect: Leveraging Social Services for Wireless Services When Network Effect Meets Congeston Effect: Leveragng Socal Servces for Wreless Servces aowen Gong School of Electrcal, Computer and Energy Engeerng Arzona State Unversty Tempe, AZ 8587, USA xgong9@asuedu

More information

Survey on Virtual Machine Placement Techniques in Cloud Computing Environment

Survey on Virtual Machine Placement Techniques in Cloud Computing Environment Survey on Vrtual Machne Placement Technques n Cloud Computng Envronment Rajeev Kumar Gupta and R. K. Paterya Department of Computer Scence & Engneerng, MANIT, Bhopal, Inda ABSTRACT In tradtonal data center

More information

Staff Paper. Farm Savings Accounts: Examining Income Variability, Eligibility, and Benefits. Brent Gloy, Eddy LaDue, and Charles Cuykendall

Staff Paper. Farm Savings Accounts: Examining Income Variability, Eligibility, and Benefits. Brent Gloy, Eddy LaDue, and Charles Cuykendall SP 2005-02 August 2005 Staff Paper Department of Appled Economcs and Management Cornell Unversty, Ithaca, New York 14853-7801 USA Farm Savngs Accounts: Examnng Income Varablty, Elgblty, and Benefts Brent

More information

Distributed Optimal Contention Window Control for Elastic Traffic in Wireless LANs

Distributed Optimal Contention Window Control for Elastic Traffic in Wireless LANs Dstrbuted Optmal Contenton Wndow Control for Elastc Traffc n Wreless LANs Yalng Yang, Jun Wang and Robn Kravets Unversty of Illnos at Urbana-Champagn { yyang8, junwang3, rhk@cs.uuc.edu} Abstract Ths paper

More information

WAN Network Design. David Tipper Graduate Telecommunications and Networking Program. Slides 10 Telcom 2110 Network Design. WAN Network Design

WAN Network Design. David Tipper Graduate Telecommunications and Networking Program. Slides 10 Telcom 2110 Network Design. WAN Network Design WAN Network Desgn Davd Tpper Graduate Telecommuncatons and Networkng Program Unversty t of Pttsburgh Sldes 10 Telcom 2110 Network Desgn WAN Network Desgn Gven Node locatons (or potental locatons) Traffc

More information

Abteilung für Stadt- und Regionalentwicklung Department of Urban and Regional Development

Abteilung für Stadt- und Regionalentwicklung Department of Urban and Regional Development Abtelung für Stadt- und Regonalentwcklung Department of Urban and Regonal Development Gunther Maer, Alexander Kaufmann The Development of Computer Networks Frst Results from a Mcroeconomc Model SRE-Dscusson

More information

"Research Note" APPLICATION OF CHARGE SIMULATION METHOD TO ELECTRIC FIELD CALCULATION IN THE POWER CABLES *

Research Note APPLICATION OF CHARGE SIMULATION METHOD TO ELECTRIC FIELD CALCULATION IN THE POWER CABLES * Iranan Journal of Scence & Technology, Transacton B, Engneerng, ol. 30, No. B6, 789-794 rnted n The Islamc Republc of Iran, 006 Shraz Unversty "Research Note" ALICATION OF CHARGE SIMULATION METHOD TO ELECTRIC

More information

The OC Curve of Attribute Acceptance Plans

The OC Curve of Attribute Acceptance Plans The OC Curve of Attrbute Acceptance Plans The Operatng Characterstc (OC) curve descrbes the probablty of acceptng a lot as a functon of the lot s qualty. Fgure 1 shows a typcal OC Curve. 10 8 6 4 1 3 4

More information

Price Competition in an Oligopoly Market with Multiple IaaS Cloud Providers

Price Competition in an Oligopoly Market with Multiple IaaS Cloud Providers Prce Competton n an Olgopoly Market wth Multple IaaS Cloud Provders Yuan Feng, Baochun L, Bo L Department of Computng, Hong Kong Polytechnc Unversty Department of Electrcal and Computer Engneerng, Unversty

More information

Fuzzy Set Approach To Asymmetrical Load Balancing In Distribution Networks

Fuzzy Set Approach To Asymmetrical Load Balancing In Distribution Networks Fuzzy Set Approach To Asymmetrcal Load Balancng n Dstrbuton Networks Goran Majstrovc Energy nsttute Hrvoje Por Zagreb, Croata goran.majstrovc@ehp.hr Slavko Krajcar Faculty of electrcal engneerng and computng

More information