Content Multi-homing: an Alternative Approach

Size: px
Start display at page:

Download "Content Multi-homing: an Alternative Approach"

Transcription

1 Content Muti-homing: an Aternative Aroach Technica Reort HKUST-CS14-01 Jason Min Wang 1, Jun Zhang 2, Brahim Bensaou 1 1 Deartment of Comuter Science and Engineering, The Hong Kong Universit of Science and Technoog 2 Network and Comuter Science Deartment, Teecom ParisTech [email protected], [email protected], [email protected] Abstract The CDN serves as an essentia eement in roviding content deiver services on the Internet; however, imited b its footrint and infuenced b variations of network conditions and user demands, no individua CDN can quaitative fufi deiver tasks antime and anwhere. As such, arge content roviders often use mutie CDNs. Nevertheess, this aroach is not on cumbersome to negotiate mutie business contracts, but aso economica inefficient, esecia for sma content roviders. A simer aternative aroach woud be to et each content rovider on sign one contract with an authoritative CDN, sa, and et hande the otentia erformance tra of individua CDN. To fufi its business commitment, coud either use its own infrastructure or rent services from other CDNs, which is sometimes indisensabe. To make an otima oerating an for oerator, we roose a new robem, caed MCDN-CM, whose objective is to minimize CDN- 0 s oerating cost. MCDN-CM is a concave minimization robem and we take advantage of the secia form of the ractica CDN-ricing function iece-wise inear concave to arrive at an otima soution through an iterative rocedure. We conduct numerica eeriments under reaistic settings and show via the eerimenta resuts that can achieve a tremendous costsaving b using our roosed agorithm. I. INTRODUCTION B moving content coser to end users, content deiver networks (CDNs) great reduce the content distribution cost and significant imrove the content access eerience. Increasing, CDNs are recognized to a a fundamenta roe in suorting the arge-scae content deiver over the Internet. Nowadas, arge content roviders (CPs) re on the CDN to deiver their content in the goba scae. For eame, YouTube uses Googe s rivate CDN to deiver videos; Netfi [1] and Huu [2] rovide their services via third-art CDNs, incuding Akamai, LimeLight, and Leve-3. The ever-growing demand for videos drives the continua u-scaing of the CDN infrastructures. From the content rovider ersective, it is desirabe that its content can be deivered to users with high quait, regardess of the geograhica ocation of the users and the ISP network the attach to. However, two facts make this requirement difficut to achieve. First, it is hard for an singe CDN rovider, even the argest one ike Akamai, to eand its footrint to ever corner of the word. Tica a CDN resents distinct erformances at different geograhica areas. Second, due to the dnamics of network conditions and the variation of server oads, the erformance of a CDN, even at the same area, fuctuates over time [1] [4]. As a consequence, content roviders ike Netfi [1] and Huu [2] use mutie Cost v r 1 v r 2 v r 3 CDN-r (a) Voume Cost 2 1 CDN-s (b) CDN-r Voume Fig. 1. (a) The iecewise inear concave ricing function of CDN-r, f r( ). (b) Motivating eame: when invoving two CDNs (CDN-r and CDN-s), the content deiver erformance gets imroved but the tota cost aid b the sma content rovider is greater than that using on CDN-r: >. CDNs to rovide high quait services; this is referred to as content muti-homing [3]. Content muti-homing enabes a content rovider to avoid the quait of eerience (QoE) tra of an singe CDN, whereas it coud resut in economic inefficienc for sma or medium content roviders. Such economic inefficienc arises from the wide-adoted charging oic of CDN: i) ou a as ou go, and ii) the more ou use, the cheaer the average rice ou get. Secifica, the CDN ricing function has a iecewise inear concave shae, as shown b Fig. 1(a). Under current charging oicies, a sma CP, when using mutie CDNs, has to a the ortions of traffic at a high charging rate for individua CDNs. This is iustrated b the eame in Fig. 1. Note that for a arge CP this robem is aeviated, since its charging voume at each CDN can easi arrive at the highest voume-rice segment, thus making the ament at a ower average rice. In addition, content muti-homing entais each CP to negotiate and sign business contracts with mutie CDN roviders, which is cumbersome. Motivation. To mitigate the above diemma, an aeaing aroach consists in etting sma CPs contact and dea with on one authoritative CDN, sa. As such, CPs no onger suffer from the negative effects of ament discount oic brought b muti-homing. It is now s resonsibiit to fufi the erformance commitment, stiuated in the service eve agreement (). To this end, must re on other CDNs in rocuring content muti-homing transarent to the CP. This new aroach is iustrated b Fig. 2. acts as a ro, aggregating demands from registered CPs and acquiring the maimum discounts from other CDNs as a giant CP. Overview. Deivering an content item to a secific area can either use s own servers or emo other CDNs. Two factors imact the decisions. One is that can not satisf the service quait requirement of certain items, restricted b its footrint and regiona service caabiities. In this case, it has

2 CP-1 CP-2... CP-N CoudFront MaCDN N CDN-K US US EU Sain HK/SG JP Austria S/A China AU Jaan US/EU/SA Brazi A/P Austraia Fig. 2. Mutie content roviders sign business contracts with one authoritative CDN with various ; the authoritative, as a customer, rents services from other CDNs, besides using its own infrastructure. to resort to other CDNs that are better imanted regiona. The other is that deivering some content items to certain areas, coud be more cost-effective in renting other CDNs services than doing it on its own network. For eame, the average cost of running one reica server in a secific area coud be so high that it hinders from running a ortion of its reica servers there. Consequent, the oerator of is insired to carefu make an oerating an: how to reaize the business commitments, in articuar the QoE requirements, b using either its own infrastructure or other CDNs services, whist aiming to minimize its oerating cost. In this aer, we he oerator make the oerating an otima, b modeing it as a concave minimization robem, referred as MCDN-CM. We sove the robem otima b invoking an iterative rocedure. Contributions. First, to the best of our knowedge, we are the first to roose the MCDN-CM robem, which has a strong motivation for the oerator of and aso benefits the content roviders, esecia smaer ones, in both rocedura and economica asects. Second, we give an agorithm to sove MCDN-CM otima. The soution can be direct transated into s oerating an. Third, this work embraces the emerging trend of CDN-interconnection [5]. Outine. Section II describes the necessar background on CDN and formuates the MCDN-CM robem. We derive the otima soution to MCDN-CM in Section III and resent the eerimenta resuts in Section IV. Fina, we concude the aer in Section V. II. CONTENT MULTI-HOMING MODEL We now resent the new content muti-homing mode. After introducing some basic terminoog and notations, we forma define the MCDN-CM robem. A. Preiminaries CDN rovider. A CDN rovider deos reica servers at strategica chosen edge ocations, aka oint of resences (PoPs), scattered across the Internet. We view each PoP as an entit that hosts servers and has a massive amount of comuting and caching ower. Different CDNs have different andscaes in terms of deoed PoPs and service caacities. We denote P the set of PoPs owned b. Content rovider. Each content rovider (e.g., YouTube) services a set of objects to users at different areas via the CDN infrastructure. Suose the set of content roviders N have signed business contracts with, shown in Fig. 2. For each CP n N, I n reresents the set of served content objects. In view of the icense issues, it is reasonabe to US : CDN S charging region, r R China : request s ocation area, a A Fig. 3. The maing from request ocation area a A to charging region r R k of two CDNs: CoudFront and MaCDN. assume that content sets of different roviders are disjoint, i.e., I n1 I n2 =, n 1 n 2. For convenience, et I = n N I n. Content object. Each content object (e.g.,, a video) ossesses various roerties [3]. For content distribution, ke roerties invove the object size, the content ouarit, and the erformance requirement. Let o i be the size of object i I. Since an object i can be requested b users in different areas, we denote A the set of geograhica areas. Let d ai be the number of times that object i has been requested b the users in area a during a eriod. Obvious, d ai refects the content ouarit for this object. Note that d ai s coud be estimated according to historica access records. Thus, we assume these arameters are avaiabe to use in our robem. The roert of erformance requirement is discussed ater. CDN ricing. Some CDN roviders (e.g., Amazon CoudFront and MaCDN) offer their cometitive rices in ubic, whereas other CDN roviders (e.g., Akamai, Limeight, and Leve3) do not ubish their ricing information. Nevertheess, the ricing oicies in the CDN market genera obe the rue that ou a as ou go and the more traffic ou use, the ower the average rice ou get. The CDN ricing has the foowing three roerties: (1) regiona-based ricing, with a rice that varies across different regions; (2) voume-based charging, where charging is based on the voume of traffic, originating from ocation areas of a secific charging region, during the biing eriod (e.g., one month); and (3) quantitative discount, where the arger the voume is consumed, the arger the reative discount is given. For eame, Amazon CoudFront divides its ricing into si charging regions [6], shown in Fig. 3. Among the si, the regiona rice of US is the cheaest; it charges 0.12$/GB for ow voume and 0.02$/GB for high voume. Let K be the set of CDNs that everages to fufi its business commitments. Let R k be the set of charging regions of CDN-k. Cear, R k1 R k2 =, k 1 k 2. For simification, et R = k K R k. Further, we denote fk r ( ), as in Fig. 1.(a), the voume ricing function of CDN-k over region r. QoE guarantee. The business contract stiuates the serviceeve-agreement () that outines the eected QoE and costs associated with the business arrangement. The meaning of QoE, for different tes of objects, varies. For eame, the QoE with resect to a Web age or image coud sim mean access atenc, whie for a YouTube video it coud ertain to the bit-rate. Instead of secifing distinct QoE metrics for diverse objects, we adot a sime unified characterization [3], the fraction of times that requests for an object can be served with sufficient QoE. We use ˆq ai to denote the erformance target set b the CP. For eame, for video i,

3 ˆq ai = 90% means that 90 ercent requests at area a shoud be served at the video s encoding rate. On the other hand, we use q ai, P and qr ai, r R k 1 to characterize the abiit (fraction of times) of PoP of and the servers in region r of CDN-k in serving requests at area a for object i with the targeted QoE, resective. To refect the QoE constraints, we define Q ai = { P, q ai ˆq ai } and Q ai = {r r R, q r ai ˆq ai}. Note that, it is ossibe that Q ai = Q ai =. In this case, the best or r wi be added. B. Cost Mode and Probem Formuation We mode PoP as a coection of reica servers used to cache content and serve cient requests. As the ufront investment, we assume has deoed M servers at PoP P. Given the demands from customers N, the CDN- 0 oerator needs to figure out how requests for objects are served. To this end, we introduce two decision variabes: i) ai, the fraction of requests at area a for object i to be directed to PoP of ; and ii) r ai, r R k, the fraction of requests to be served b servers in region r of CDN-k. To suort { ai }, we assume m servers have to be rovisioned at PoP. The goa is to minimize s oerating cost that consists of two arts: i) the running cost of its infrastructure, {m } servers; ii) the cost aid to other CDNs in K. Running cost of P. In ractice, the running cost of PoP deends on severa factors, ike eectricit, bandwidth, and ta. Cost ike eectricit even varies temora. To simif, we use h to reresent the average hosting cost er server at PoP. Then the running cost of PoP becomes m h. Abeit sime, it has catured the fundamenta asects of CDN s oerating cost, i.e., the cost increases with m ; h varies across different areas. Further, we use b to denote the average rocessing ower of one server at PoP ; b is normaized to be the average number of requests each server can resond to. Note that this simified mode was used in [7], [8] as we. Cost aid to other CDNs. The ortions of requests redirected to other CDNs aggregate to be the charging voume that CDN- 0, as the customer, as to other CDNs. The cost aid to CDNk is r R k fk r( a A i I d aio i r ai ). Probem formuation. Given the above cost mode, the goa is to minimize the tota oerating cost of. We ca this the Mutie CDNs Cost Minimization (MCDN-CM) robem, defined as: min,m s.t. k,r f r k ( d ai o i r ) ai + m h (1) ai + r ai = 1, a A, i I (2) Q ai r Q ai ai 0, a A, i I, Q ai (3) r ai 0, a A, i I, r Q ai (4) r ai = 0, a A, i I, r / Q ai (5) d ai ai m b, P (6) 0 m M, P (7) 1 In ractice, {q ai } can be estimated based on the access ogs of ; {qai r } can be estimated from the that has signed with CDN-k. (1) is s oerating cost. (2) ensures that an user request for object i at area a is served. The QoE requirements are catured via Q ai and Q ai, encasuated in (3), (4), and (5). (6) states that the eected tota oad of servers in PoP shoud not eceed the service caacit. (7) imits the number of active servers of PoP. Athough m is integra in essence, we ma rea it to be continuous in MCDN-CM, in view of: i) a PoP coud comrise hundreds of servers; and ii) b and h er se are rough estimates. Aart from the service caacit constraint (6), we coud aso add constraints ike the bandwidth caacit constraint, d aio i ai m u. The soution discussed hereinafter sti works. MCDN-CM is a concave minimization robem [9]. Since decision variabes are we bounded and constraints are inear, the feasibe set D is a bounded ohedron. We assume D is non-emt. B Weierstrass Theorem, MCDN-CM has a goba otima soution, which is attained at an etreme oint of D [9]. However, instead of using a genera aroach to sove a concave minimization robem, we wi take advantage of the secia structure of fk r ( ) iecewise inear concave. C. Discussion Liu et a. studied content muti-homing [3] from the ersective of arge content roviders. Their aroach requires each content rovider to sign mutie CDN contracts and is not economica efficient for sma content roviders. In this aer, we advocate the secia roe of the authoritative CDN- 0. Thus, the muti-homing robem in [3] and MCDN-CM differ in the articiation of, with a different objective function and additiona constraints. Athough Liu et a. gave a secia and efficient agorithm in soving the muti-homing robem, we cannot use their aroach to sove MCDN-CM. This is because the etra server caacit constraints (5) (or the otentia bandwidth caacit constraints) rohibit transforming MCDN-CM into the so-caed assignment robem [3] in the first ste, thus making the aroach in [3] infeasibe. III. GLOBAL OPTIMIZATION To reach an otima soution to MCDN-CM, we transform the concave minimization robem into a biinear rogram b eoiting the secia structure of the CDN ricing function and sove the biinear rogram in an iterative manner. A. Eiminating m Coser observation to MCDN-CM reveas that the constraint (7) must be tight in the otima soution (, m ). That is, d ai ai = m b, P. With this roert, we can eiminate the decision variabe m from MCDN-CM, and get an equivaent robem MCDM-CM-E: min f r ( k d ai o i r ) ai + c d ai ai (8) k,r s.t. (2), (3), (4), (5) d ai ai s, P (9) where c = h /b and s = M b. Note that c can be viewed as the normaized er request service cost in PoP and s is the maimum service caacit of PoP.

4 B. Biinear Transformation Tabe I demonstrates the concrete ricing functions of CoudFront and MaCDN. The ricing function is iece-wise inear concave (see Fig. 1(a)). Secifica, fk r ( ) can be written in the foowing form u 1 r r + e 1 r, r [vr 0 = 0, vr) 1 fk r u 2 ( r ) = r r + e 2 r, r [vr, 1 vr) 2 (10) u nr r r + e nr r, r [vr nr 1, ) where u 1 r > u 2 r > > u nr r. n r is the number of charging intervas and u r(1 n r ) refects the average rice (er GB) in the voume interva [vr 1, vr). Equivaent, f r k ( r ) = min {u r r + e r }. (11) 1 n r For brevit, et r = d aio i r ai reresent the charging voume of region r R. Given the voume interva in which r fas, fk r ( ) turns into a inear function, whie MCDF-CM- E becomes a inear rogram. For this urose, we introduce for each region r a set of binar variabes r (1 n r ), indicating whether r fas in [vr 1, vr). That is, { 1, if r r [vr 1, v = r) (12) 0, otherwise. We reformuate the robem MCDN-CM-E: i) modif the objective function (8) using binar variabes r, fk r( r) = [u r r + e r]r; ii) add etra constraints, r = 1, v 1 r r r v rr, and r = {0, 1}. Reaing the integer constraint of rs eads to the foowing robem, referred to as MCDN-CM-ER, r( u r r + e ) r + c d ai ai (13) min, k,r s.t. (2), (3), (4), (5), (9) vr 1 r r vr r, r R (14) r = 1, r R (15) r 0, r R, 1 n r (16) Athough we rea the integer constraint of r, Theorem 1 (see roof in Aendi) estabishes the equivaence of MCDN- CM-ER and MCDN-CM-E. Therefore, we devote to soving MCDN-CM-ER instead. Theorem 1: (, ) is an otima soution to MCDN-CM- ER, if and on if is an otima soution to MCDN-CM-E. C. Agorithm for Soving MCDN-CM-ER MCDN-CM-ER is a biinear rogram with joint constrained feasibe region. Aong the ine of Dnamic Cost Udating Procedure (DCUP) [10], we resent an eact and continuous agorithm, Ag. 1, to sove MCDN-CM-ER. DCUP rocedure is based on the observation that the feasibe sets of variabes { r ai } and { r} are joint constrained on b constraint (14). If we fi one te of variabes, the remaining Agorithm 1 DCUP for MCDN-CM-ER 1: Initiaization: et r (0) be the initia vector of { r} where 1 r(0) = 1 and r(0) = 0, 1; et t 1; 2: Iteration-t: 2.1: Sove MCDN-LP ( (t 1) ) : (t) arg min { MCDN-LP ( (t 1) )} ; 2.2: Sove MCDN-LP ( (t) ) : (t) arg min { MCDN-LP ( (t) )} ; 3: If (t) = (t 1), then sto; otherwise, et t t + 1 and go to Ste 2; robem becomes an otimization with resect to the other te of variabes and consists in a inear rogram. Secifica, b fiing, we get the foowing inear robem MCDN-LP(), min r u rr + c d ai ai k,r (17) s.t. (2), (3), (4), (5), (9) Simiar, b fiing the variabe, we get the other inear robem MCDN-LP(), ) min r( u r r + e r k,r (18) s.t. (14), (15), (16) Ag. 1 roceeds b soving two (coued) inear rograms iterative. It is worth noting that MCDN-LP() itsef does not incude the constraint (14). But when Ag. 1 terminates, the constraint (14) wi be satisfied, because this constraint aread aears in MCDN-LP(). Theorem 2 estabishes the correctness of Ag. 1 in soving MCDN-CM-ER and Theorem 3 estabishes the convergence of Ag. 1 (see roof in Aendi). Theorem 2: The fina soution (, ỹ) given b Ag. 1 is an otima soution to MCDN-CM-ER. Theorem 3: The Ag. 1 stos in a finite number of iterations. IV. PERFORMANCE EVALUATION MCDN-CM-ER aims to minimize the oerating cost of, whist fufiing its business commitment, that requests from cients for content rovided b registered CPs shoud be satisfied with sufficient QoE secified in the s. It is natura to observe the cost achieved from different oerating ans. A. Eerimenta Setu Datasets. To make the numerica eeriments as reaistic as ossibe, we used the Youtube Dataset rovided b [11]. We ran the eeriments on 5 grous of datasets. Since simiar resuts were observed, we on reort resuts of one grou for brevit. Each grou dataset consists of 5 das records. We random distribute them to N = 10 content roviders (CPs). Tica, a 5-da dataset corresonds to severa thousands of TB traffic, which we beieve is a medium size 2 and is aroriate for our eeriments. Imortant fieds of the dataset 2 The mathematica mode of mutie-cdn cost minimization (MCDN-CM) er se does not deend on the fideit of an inut arameters.

5 TABLE I. REGIONAL PRICING FUNCTIONS USED IN THE EXPERIMENTS CDN Region [0, 10TB) [10TB, 50TB) [50TB, 150TB) [150TB, 500TB) [500TB, 1PB) [1PB, 5PB) [5PB, ) US $0.12 /GB $0.08 /GB $0.06 /GB $0.04 /GB $0.03 /GB $0.025 /GB $0.02 /GB EU $0.12 /GB $0.08 /GB $0.06 /GB $0.04 /GB $0.03 /GB $0.025 /GB $0.02 /GB CoudFront HK/SG $0.19 /GB $0.14 /GB $0.12 /GB $0.10 /GB $0.08 /GB $0.07 /GB $0.06 /GB JP $0.201 /GB $0.148 /GB $0.127 /GB $0.106 /GB $0.085 /GB $0.075 /GB $0.065 /GB S/A $0.25 /GB $0.20 /GB $0.18 /GB $0.16 /GB $0.14 /GB $0.13 /GB $0.125 /GB AU $0.19 /GB $0.14 /GB $0.12 /GB $0.11 /GB $0.095 /GB $0.090 /GB $0.085 /GB MaCDN US/EU/SA $0.07 /GB $0.06 /GB $0.05 /GB $0.04 /GB $0.035 /GB $0.03 /GB $0.02 /GB A/P $0.10 /GB $0.074 /GB $0.064 /GB $0.053 /GB $0.043 /GB $0.037 /GB $0.032 /GB invove the video id, the views, the ength, and the video rate (KB/s). The roduct of ength and rate ieds the video size. Based on the rate vaue, the videos are categorized in two tes: ow bit-rate and high bit-rate. We distinguish these two tes, as each te has different QoE requirements. Pricing and QoE settings. We deiberate choose CoudFront () and MaCDN () to form the set K, as their regiona ricing is ubic avaiabe, thereb eading to a reaistic setting of ricing functions fk r ( ). The concrete ricing function is iustrated in Tabe I. On average, the ricing of MaCDN is ower than that of CoudFront. Liu et a. in [3] conducted eeriments via PanetLab and measured the CDN s erformance in servicing requests from 7 ocation areas (as are shown in Fig. 3). To make the settings of QoE arameters {qai r } reaistic, we reuse their coected data and set A to hod those 7 areas as we. The detaied information on QoE arameter settings is summarized in Tabe II. In genera, CoudFront resents a high quait in servicing both ow bit-rate and high bit-rate videos, whie MaCDN resents varing erformances in serving different tes of videos at distinct areas. Fina, the QoE target q ai is set to 90%, which determines the vaues of Q ai and Q ai in turn. Setting of. To mode the footrint of, we assume 3 PoPs have been deoed at each area in A. For eame, P 1, P 2, and P 3 are three PoPs to serve requests from the US area. In tota contains P = 21 PoPs. We assume that three PoPs at the same area resent simiar erformance, i.e., q 1 ai = q2 ai = q3 ai. The concrete vaues are given in Tabe II. Furthermore, the genera cost of running an active server, h, shoud refect the regiona cost variations. To this end, we set the vaues of {h }, referring to the ricing for Virtua Machines of Amazon EC2 [12]. The setting of {h } is summarized in Tabe III, where different grous corresond to different eves of hosting rices. Last, we set M, the quota of servers in, between [20, 35] random. B. Evauation Methodoog The soution given b Ag. 1, referred as OPT, offers the detaied oerating an of. An feasibe oerating an invoves two asects: i) PoP activation, that m servers of PoP wi be activated; and ii) request assignments 3, that r ai (r R k) ortion of requests for object i at area a wi be directed to region r of CDN-k and ai ortion wi be redirected to PoP of. Together {m }, { r ai }, { ai } determine the oerating cost. To evauate the erformance of 3 The requests redirection ma either use eisting DNS-based aroach or emo the emerging technique of software defined networking (SDN). TABLE II. MEASURED CDN PERFORMANCE SETTINGS US Sain Austria China Jaan Brazi Austraia 99 a b a The first row corresonds to the erformance in servicing ow bit-rate videos. b The second row corresonds to the erformance in servicing high bit-rate videos. TABLE III. AVERAGE HOSTING PRICE OF SERVER IN POP OF h grou US Sain Austria China Jaan Brazi Austraia Grou c Grou Grou Grou c in US doars er month. Costs (Kio US Doar) OPT QoE-on Random Greed Grous of { h } Fig. 4. The gross oerating cost of using different methods, with varing grous of {h }. Ag. 1 in minimizing the oerating cost, we comare it with three aternatives: i) QoE-on, which awas choose the region r or PoP that can rovide the best QoE (maimum qai r or q q ai ); ii) random, which icks an one among those regions or PoPs that satisf the QoE requirements; and iii) greed, which seects from the QoE-satisfied candidates the one that eads to the owest cost after current assignment. Note that the above three aternatives a roceed b making assignments for requests sequentia in a uniform random order.

6 Traffic (TB) OPT QoE-on Random Greed Traffic (TB) Grou-1 Grou-2 Grou-3 Grou-4 Fig. 5. Traffic distribution of four methods under Grou-1 {h }. Fig. 6. Traffic distribution using OPT, under four grous of {h }. We use the Pthon interface of Gurobi-5.1 (under the academic icense) to sove MCDN-LP(). With around thousands of ocation-objects, Gurobi-5.1 can sove MCDN- LP() in about 10 minutes under a machine with 4 AMD Oteron 844 (1.8GHz) CPU and 8GB RAM. In a our eeriments, Ag. 1 terminates with ess than 3 iterations. C. Eerimenta Resuts First, we observe the cost aid b for accomishing its business commitment with fu QoE grantees. We comare the vaues of the oerating cost, resuting from the four different agorithms mentioned above. Fig. 4 shows the resuts with varing grous of {h }. As eected, OPT eads to the east month cost. In genera, the greed method erforms much better than the other two aternatives: QoE-on and random. This is because, these two methods do not take into consideration at a the cost when making request assignments. Abeit, the curve of greed seems to be cose to that of OPT, this atter sti brings a considerabe saving. For eame, under this articuar arameter settings, OPT saves , , , and US doars for each month under the four grous of h searate, comared with greed. Net, we dissect how ortions of traffic are distributed using different methods. From Fig. 5, we can see that OPT rioritize using s own infrastructure under Grou- 1 {h }, since Grou-1 s hosting rices are reative sma. Further, OPT refers to, since the average rice of is ower than that of in genera. But it sti has to seek he from, since cannot satisf the erformance target in areas ike China, Brazi, and Austraia. This oint is further refected b the traffic distribution of greed. From Tabe II, we observe that rovides the best erformance across a the areas, which arises from the far-fung footrint of CoudFront as a goba CDN rovider. B virtue of this fact, QoE-on heavi reies on. On the other hand, random equa uses the three CDNs and wi suffer from the negative effects of using mutie CDNs caused b the ament discount oic, when the tota traffic is of medium size, as in this case. Fina, we see how the traffic distribution changes when the average hosting rice of increases. As shown in Fig. 6, when the hosting costs of PoPs increase, OPT begins Fig. 7. Traffic (TB) Grou-1 Grou-2 Grou-3 Grou-4 Traffic distribution using greed, under four grous of {h }. to re more on other CDNs, articuar, in servicing requests. The greed aroach refects such change as we, but not as evident as does OPT, as shown in Fig. 7. V. CONCLUSION This aer is motivated b the content muti-homing robem in [3]. Content muti-homing is economica inefficient, articuar for the mriad of sma content roviders. We roose an aternative aroach where each content rovider on signs the contract with the authoritative, and takes charge of handing the otentia QoE tra of an singe CDN, b utiizing other CDNs to fufi its business commitment. We identif the MCDN-CM robem that hes the oerator of make an otima oerating an in terms of minimizing its oerating cost. Under reaistic settings, the eerimenta resuts show that our agorithm for soving MCDN-CM resuts in considerabe amount of mone saving. In the future work, we an to stud the imacts of renting other CDNs services on defining the ricing an of CDN- 0. For eame, can raise its rice without oosing otentia customers? VI. ACKNOWLEDGMENTS This work was in art suorted b UGC grant FS- GRF14EG24.

7 REFERENCES [1] V. Adhikari, Y. Guo, F. Hao, M. Varveo, V. Hit, M. Steiner, and Z.- L. Zhang, Unreeing netfi: Understanding and imroving muti-cdn movie deiver, in INFOCOM, 2012 Proceedings IEEE, March 2012, [2] V. Adhikari, Y. Guo, F. Hao, V. Hit, and Z.-L. Zhang, A tae of three cdns: An active measurement stud of huu and its cdns, in Comuter Communications Workshos (INFOCOM WKSHPS), 2012 IEEE Conference on, March 2012, [3] H. H. Liu, Y. Wang, Y. R. Yang, H. Wang, and C. Tian, Otimizing cost and erformance for content mutihoming, in Proceedings of the ACM SIGCOMM 2012 Conference on Aications, Technoogies, Architectures, and Protocos for Comuter Communication, ser. SIGCOMM 12. New York, NY, USA: ACM, 2012, [4] A.-J. Su, D. R. Choffnes, A. Kuzmanovic, and F. E. Bustamante, Drafting behind akamai: inferring network conditions based on cdn redirections, IEEE/ACM Trans. Netw., vo. 17, no. 6, , Dec [5] N. B. Ben Niven-Jenkins, Francois Le Faucheur, Content distribution network interconnection (cdni) robem statemen-rfc6707, htt://datatracker.ietf.org/doc/draft-ietf-cdni-robem-statement/, [6] Amazon. Amazon coudfront ricing. htt://aws.amazon.com/coudfront/ricing/. [7] P. X. Gao, A. R. Curtis, B. Wong, and S. Keshav, It s not eas being green, in Proceedings of the ACM SIGCOMM 2012 Conference on Aications, Technoogies, Architectures, and Protocos for Comuter Communication, ser. SIGCOMM 12. New York, NY, USA: ACM, 2012, [8] Q. Zhang, Q. Zhu, M. F. Zhani, and R. Boutaba, Dnamic service acement in geograhica distributed couds, in Proceedings of the 2012 IEEE 32nd Internationa Conference on Distributed Comuting Sstems, ser. ICDCS 12, 2012, [9] R. Horst and H. Tu, Goba Otimization: Deterministic Aroaches. Sringer, [10] P. M. Nahaetan, A. Pardaos, A biinear reaation based agorithm for concave iecewise inear network fow robems, Journa of Industria and Management Otimization, vo. 3, no. 1, , [11] X. Cheng, C. Dae, and J. Liu, Dataset for statistics and socia network of outube videos, Simon Fraster Universit, Canada. htt://netsg.cs.sfu.ca/outubedata/, [12] Amazon. Amazon ec2 ricing. htt://aws.amazon.com/ec2/ricing/. APPENDIX Proof of Theorem 1: To rove the theorem, we wi show that an otima soution to one of the two robems is a feasibe soution to the other with identica objective vaue. Suose is an otima soution to MCDN-CM-E. We ma construct the vector ỹ according to (12). That is, for each r R, if r [vr 1, vr), et ỹr = 1; and, 1 n r, et ỹr = 0. Since satisfies constraints (2), (3), (4), (5), and (9), and ỹ satisfies constraints (14), (15), and (16), (, ỹ) is a feasibe soution to MCDN-CM-ER. Further, fk r( r) = u r r + e r, where r [vr 1, vr). Thus, the objective vaue of MCDN- CM-ER given b the feasibe soution (, ỹ) is identica to the otima objective vaue of MCDN-CM-E. Net, we show the oosite direction. Suose (, ỹ) is an otima soution to MCDN-CM-ER. First, it is eas to check that is aso feasibe to MCDN-CM-E. Second, we need to show that the objective vaues of the two robems are identica. Note that the objective functions of the two (8) and (13) differs on in the first art. Thus, we on need to check whether fk r( r) equas r( ỹ u r r + er). Or ut in another wa, we eamine whether ỹr resects (12). B fiing the vaue of vector as, the MCDN-CM-ER robem reduces to a inear rogram with resect to, that is, the robem LP( ). Note that LP( ) can be fu searated into R subrobems for each r R. Hence, in the foowing, we on focus on a secific region r. Since (, ỹ) is an otima soution to MCDN-CM-ER, ỹ must be the otima soution to LP( ) as we. On the other hand, we assume that r [vr 1, vr]. That is, in the otima soution, the vaue r ies in the segment indicated b. Reca (11) and thus fr( r ) = min {u r r + e r} = u r r + e r. Based on this, we 1 n r construct a new vaue ŷ r of vector r : et ŷr = 1; and for an, et ŷr = 0. Then the new constructed vector ŷ r satisfies: i) the constraint (14), because r [vr 1, vr]; and ii) ŷ r = arg min{ r(u r r + e r) r = 1, r 0}. Hence, ŷ r is an otima soution to the subrobem of LP( ) with resect to r. A simiar resut hod for an other r R. Therefore, ŷ is an otima soution to LP( ). Since ỹ is aso an otima soution to LP( ), we concude (, ŷ) is aso an otima soution to MCDN-CM-ER 3. These two facts: i) fr( r ) = r( ŷ u r r + er), according to how we construct ) = ỹ r( u r r + e r), because ŷ r ; and ii) r( ŷ u r r + e r both and ŷ are otima to LP( ), aow us to concude that fk r( r) is equa to r( ỹ u r r + er). So we finish roving that the objective vaue of MCDN-CM-E given b equas the objective vaue of the otima soution (, ỹ) of MCDN- CM-ER. Proof of Theorem 2: Let (, ỹ) be the returned soution of Ag. 1. For notationa simification, et g(, ) reresent the objective function (13) of MCDN-CM-ER and et D be the feasibe region. B virtue of the stoing condition in Ste 3 of Ag. 1, we get: (i) = arg min{mcdn-lp(ỹ)}; and (ii) ỹ = arg min{mcdn-lp( )}. (ii) eads to ỹ = arg min g(, ). Since ỹ satisfies the constraint (14) aread (even though (14) is miss- (,) D ing in MCDN-LP(ỹ)), combined with (i), we get = arg min (,ỹ) D g(, ỹ). The above two equations show that (, ỹ) = arg min g(, ). Therefore, the soution (, ỹ) (,) D returned b Ag. 1 is an otima soution of MCDN-CM-ER. Proof of Theorem 3: Theorem 3 can be roven in the same aroach as Theorem-6 in [10]. 3 Note that in this roof, we don t care whether ỹ = ŷ.

TOPOLOGICAL DESIGN OF MULTIPLE VPNS OVER MPLS NETWORK Anotai Srikitja David Tipper

TOPOLOGICAL DESIGN OF MULTIPLE VPNS OVER MPLS NETWORK Anotai Srikitja David Tipper TOPOLOGICAL DESIGN OF MULTIPLE VPNS OVER MPLS NETWORK Anotai Sriitja David Tier Det. of Information Science and Teecommunications University of Pittsburgh N. Beefied Avenue, Pittsburgh, PA 60 ABSTRACT

More information

ILP Formulation and K-Shortest Path Heuristic for the RWA Problem with Allocation of Wavelength Converters

ILP Formulation and K-Shortest Path Heuristic for the RWA Problem with Allocation of Wavelength Converters ILP Formuation and K-Shortest Path Heuristic for the RWA Probem with Aocation of Waveength Converters Karcius D.R. Assis 1, A. C. B. Soares, Wiiam F. Giozza and Heio Wadman 1 UEFS- State University of

More information

Secure Network Coding with a Cost Criterion

Secure Network Coding with a Cost Criterion Secure Network Coding with a Cost Criterion Jianong Tan, Murie Médard Laboratory for Information and Decision Systems Massachusetts Institute of Technoogy Cambridge, MA 0239, USA E-mai: {jianong, medard}@mit.edu

More information

Australian Bureau of Statistics Management of Business Providers

Australian Bureau of Statistics Management of Business Providers Purpose Austraian Bureau of Statistics Management of Business Providers 1 The principa objective of the Austraian Bureau of Statistics (ABS) in respect of business providers is to impose the owest oad

More information

Teamwork. Abstract. 2.1 Overview

Teamwork. Abstract. 2.1 Overview 2 Teamwork Abstract This chapter presents one of the basic eements of software projects teamwork. It addresses how to buid teams in a way that promotes team members accountabiity and responsibiity, and

More information

SIMPLE MODELS OF TRANSMISSION LINE IN MATHCAD ENVIROMENTS PROSTE MODELE LINII DŁUGIEJ W ŚRODOWISKU MATHCAD

SIMPLE MODELS OF TRANSMISSION LINE IN MATHCAD ENVIROMENTS PROSTE MODELE LINII DŁUGIEJ W ŚRODOWISKU MATHCAD ELEKTRYKA Zeszyt () Ro LVIII Piotr JANKOWSKI Deartment of Marine Eectrica Power Engineering, Gdynia Maritime University SIMPLE MODELS OF TRANSMISSION LINE IN MATHCAD ENVIROMENTS Summary. The artice resents

More information

Betting Strategies, Market Selection, and the Wisdom of Crowds

Betting Strategies, Market Selection, and the Wisdom of Crowds Betting Strategies, Market Seection, and the Wisdom of Crowds Wiemien Kets Northwestern University [email protected] David M. Pennock Microsoft Research New York City [email protected]

More information

Simultaneous Routing and Power Allocation in CDMA Wireless Data Networks

Simultaneous Routing and Power Allocation in CDMA Wireless Data Networks Simutaneous Routing and Power Aocation in CDMA Wireess Data Networks Mikae Johansson *,LinXiao and Stephen Boyd * Department of Signas, Sensors and Systems Roya Institute of Technoogy, SE 00 Stockhom,

More information

CERTIFICATE COURSE ON CLIMATE CHANGE AND SUSTAINABILITY. Course Offered By: Indian Environmental Society

CERTIFICATE COURSE ON CLIMATE CHANGE AND SUSTAINABILITY. Course Offered By: Indian Environmental Society CERTIFICATE COURSE ON CLIMATE CHANGE AND SUSTAINABILITY Course Offered By: Indian Environmenta Society INTRODUCTION The Indian Environmenta Society (IES) a dynamic and fexibe organization with a goba vision

More information

TERM INSURANCE CALCULATION ILLUSTRATED. This is the U.S. Social Security Life Table, based on year 2007.

TERM INSURANCE CALCULATION ILLUSTRATED. This is the U.S. Social Security Life Table, based on year 2007. This is the U.S. Socia Security Life Tabe, based on year 2007. This is avaiabe at http://www.ssa.gov/oact/stats/tabe4c6.htm. The ife eperiences of maes and femaes are different, and we usuay do separate

More information

Order-to-Cash Processes

Order-to-Cash Processes TMI170 ING info pat 2:Info pat.qxt 01/12/2008 09:25 Page 1 Section Two: Order-to-Cash Processes Gregory Cronie, Head Saes, Payments and Cash Management, ING O rder-to-cash and purchase-topay processes

More information

Face Hallucination and Recognition

Face Hallucination and Recognition Face Haucination and Recognition Xiaogang Wang and Xiaoou Tang Department of Information Engineering, The Chinese University of Hong Kong {xgwang1, xtang}@ie.cuhk.edu.hk http://mmab.ie.cuhk.edu.hk Abstract.

More information

Fast Robust Hashing. ) [7] will be re-mapped (and therefore discarded), due to the load-balancing property of hashing.

Fast Robust Hashing. ) [7] will be re-mapped (and therefore discarded), due to the load-balancing property of hashing. Fast Robust Hashing Manue Urueña, David Larrabeiti and Pabo Serrano Universidad Caros III de Madrid E-89 Leganés (Madrid), Spain Emai: {muruenya,darra,pabo}@it.uc3m.es Abstract As statefu fow-aware services

More information

A Virtual Machine Dynamic Migration Scheduling Model Based on MBFD Algorithm

A Virtual Machine Dynamic Migration Scheduling Model Based on MBFD Algorithm International Journal of Comuter Theory and Engineering, Vol. 7, No. 4, August 2015 A Virtual Machine Dynamic Migration Scheduling Model Based on MBFD Algorithm Xin Lu and Zhuanzhuan Zhang Abstract This

More information

CONTRIBUTION OF INTERNAL AUDITING IN THE VALUE OF A NURSING UNIT WITHIN THREE YEARS

CONTRIBUTION OF INTERNAL AUDITING IN THE VALUE OF A NURSING UNIT WITHIN THREE YEARS Dehi Business Review X Vo. 4, No. 2, Juy - December 2003 CONTRIBUTION OF INTERNAL AUDITING IN THE VALUE OF A NURSING UNIT WITHIN THREE YEARS John N.. Var arvatsouakis atsouakis DURING the present time,

More information

Load Balancing in Distributed Web Server Systems with Partial Document Replication *

Load Balancing in Distributed Web Server Systems with Partial Document Replication * Load Baancing in Distributed Web Server Systems with Partia Document Repication * Ling Zhuo Cho-Li Wang Francis C. M. Lau Department of Computer Science and Information Systems The University of Hong Kong

More information

Chapter 3: e-business Integration Patterns

Chapter 3: e-business Integration Patterns Chapter 3: e-business Integration Patterns Page 1 of 9 Chapter 3: e-business Integration Patterns "Consistency is the ast refuge of the unimaginative." Oscar Wide In This Chapter What Are Integration Patterns?

More information

Cooperative Content Distribution and Traffic Engineering in an ISP Network

Cooperative Content Distribution and Traffic Engineering in an ISP Network Cooperative Content Distribution and Traffic Engineering in an ISP Network Wenjie Jiang, Rui Zhang-Shen, Jennifer Rexford, Mung Chiang Department of Computer Science, and Department of Eectrica Engineering

More information

Comparison of Traditional and Open-Access Appointment Scheduling for Exponentially Distributed Service Time

Comparison of Traditional and Open-Access Appointment Scheduling for Exponentially Distributed Service Time Journa of Heathcare Engineering Vo. 6 No. 3 Page 34 376 34 Comparison of Traditiona and Open-Access Appointment Scheduing for Exponentiay Distributed Service Chongjun Yan, PhD; Jiafu Tang *, PhD; Bowen

More information

Pay-on-delivery investing

Pay-on-delivery investing Pay-on-deivery investing EVOLVE INVESTment range 1 EVOLVE INVESTMENT RANGE EVOLVE INVESTMENT RANGE 2 Picture a word where you ony pay a company once they have deivered Imagine striking oi first, before

More information

Virtual trunk simulation

Virtual trunk simulation Virtua trunk simuation Samui Aato * Laboratory of Teecommunications Technoogy Hesinki University of Technoogy Sivia Giordano Laboratoire de Reseaux de Communication Ecoe Poytechnique Federae de Lausanne

More information

Life Contingencies Study Note for CAS Exam S. Tom Struppeck

Life Contingencies Study Note for CAS Exam S. Tom Struppeck Life Contingencies Study Note for CAS Eam S Tom Struppeck (Revised 9/19/2015) Introduction Life contingencies is a term used to describe surviva modes for human ives and resuting cash fows that start or

More information

Discounted Cash Flow Analysis (aka Engineering Economy)

Discounted Cash Flow Analysis (aka Engineering Economy) Discounted Cash Fow Anaysis (aka Engineering Economy) Objective: To provide economic comparison of benefits and costs that occur over time Assumptions: Future benefits and costs can be predicted A Benefits,

More information

Finance 360 Problem Set #6 Solutions

Finance 360 Problem Set #6 Solutions Finance 360 Probem Set #6 Soutions 1) Suppose that you are the manager of an opera house. You have a constant margina cost of production equa to $50 (i.e. each additiona person in the theatre raises your

More information

Betting on the Real Line

Betting on the Real Line Betting on the Rea Line Xi Gao 1, Yiing Chen 1,, and David M. Pennock 2 1 Harvard University, {xagao,yiing}@eecs.harvard.edu 2 Yahoo! Research, [email protected] Abstract. We study the probem of designing

More information

Advanced ColdFusion 4.0 Application Development - 3 - Server Clustering Using Bright Tiger

Advanced ColdFusion 4.0 Application Development - 3 - Server Clustering Using Bright Tiger Advanced CodFusion 4.0 Appication Deveopment - CH 3 - Server Custering Using Bri.. Page 1 of 7 [Figures are not incuded in this sampe chapter] Advanced CodFusion 4.0 Appication Deveopment - 3 - Server

More information

Pricing and Revenue Sharing Strategies for Internet Service Providers

Pricing and Revenue Sharing Strategies for Internet Service Providers Pricing and Revenue Sharing Strategies for Internet Service Providers Linhai He and Jean Warand Department of Eectrica Engineering and Computer Sciences University of Caifornia at Berkeey {inhai,wr}@eecs.berkeey.edu

More information

Normalization of Database Tables. Functional Dependency. Examples of Functional Dependencies: So Now what is Normalization? Transitive Dependencies

Normalization of Database Tables. Functional Dependency. Examples of Functional Dependencies: So Now what is Normalization? Transitive Dependencies ISM 602 Dr. Hamid Nemati Objectives The idea Dependencies Attributes and Design Understand concepts normaization (Higher-Leve Norma Forms) Learn how to normaize tabes Understand normaization and database

More information

A Simple Model of Pricing, Markups and Market. Power Under Demand Fluctuations

A Simple Model of Pricing, Markups and Market. Power Under Demand Fluctuations A Simle Model of Pricing, Markus and Market Power Under Demand Fluctuations Stanley S. Reynolds Deartment of Economics; University of Arizona; Tucson, AZ 85721 Bart J. Wilson Economic Science Laboratory;

More information

Multiperiod Portfolio Optimization with General Transaction Costs

Multiperiod Portfolio Optimization with General Transaction Costs Multieriod Portfolio Otimization with General Transaction Costs Victor DeMiguel Deartment of Management Science and Oerations, London Business School, London NW1 4SA, UK, [email protected] Xiaoling Mei

More information

Ricoh Legal. ediscovery and Document Solutions. Powerful document services provide your best defense.

Ricoh Legal. ediscovery and Document Solutions. Powerful document services provide your best defense. Ricoh Lega ediscovery and Document Soutions Powerfu document services provide your best defense. Our peope have aways been at the heart of our vaue proposition: our experience in your industry, commitment

More information

A Similarity Search Scheme over Encrypted Cloud Images based on Secure Transformation

A Similarity Search Scheme over Encrypted Cloud Images based on Secure Transformation A Simiarity Search Scheme over Encrypted Coud Images based on Secure Transormation Zhihua Xia, Yi Zhu, Xingming Sun, and Jin Wang Jiangsu Engineering Center o Network Monitoring, Nanjing University o Inormation

More information

Multi-Robot Task Scheduling

Multi-Robot Task Scheduling Proc of IEEE Internationa Conference on Robotics and Automation, Karsruhe, Germany, 013 Muti-Robot Tas Scheduing Yu Zhang and Lynne E Parer Abstract The scheduing probem has been studied extensivey in

More information

GREEN: An Active Queue Management Algorithm for a Self Managed Internet

GREEN: An Active Queue Management Algorithm for a Self Managed Internet : An Active Queue Management Agorithm for a Sef Managed Internet Bartek Wydrowski and Moshe Zukerman ARC Specia Research Centre for Utra-Broadband Information Networks, EEE Department, The University of

More information

Accreditation: Supporting the Delivery of Health and Social Care

Accreditation: Supporting the Delivery of Health and Social Care Accreditation: Supporting the Deivery of Heath and Socia Care PHARMACY E F P T O L P E D P E C M F D T G L E F R Accreditation: Supporting the Deivery of Heath and Socia Care June 9, 2015 marks Word Accreditation

More information

Ricoh Healthcare. Process Optimized. Healthcare Simplified.

Ricoh Healthcare. Process Optimized. Healthcare Simplified. Ricoh Heathcare Process Optimized. Heathcare Simpified. Rather than a destination that concudes with the eimination of a paper, the Paperess Maturity Roadmap is a continuous journey to strategicay remove

More information

Wide-Area Traffic Management for. Cloud Services

Wide-Area Traffic Management for. Cloud Services Wide-Area Traffic Management for Coud Services Joe Wenjie Jiang A Dissertation Presented to the Facuty of Princeton University in Candidacy for the Degree of Doctor of Phiosophy Recommended for Acceptance

More information

Vendor Performance Measurement Using Fuzzy Logic Controller

Vendor Performance Measurement Using Fuzzy Logic Controller The Journa of Mathematics and Computer Science Avaiabe onine at http://www.tjmcs.com The Journa of Mathematics and Computer Science Vo.2 No.2 (2011) 311-318 Performance Measurement Using Fuzzy Logic Controer

More information

Service Network Design with Asset Management: Formulations and Comparative Analyzes

Service Network Design with Asset Management: Formulations and Comparative Analyzes Service Network Design with Asset Management: Formulations and Comarative Analyzes Jardar Andersen Teodor Gabriel Crainic Marielle Christiansen October 2007 CIRRELT-2007-40 Service Network Design with

More information

On Multicast Capacity and Delay in Cognitive Radio Mobile Ad-hoc Networks

On Multicast Capacity and Delay in Cognitive Radio Mobile Ad-hoc Networks On Multicast Caacity and Delay in Cognitive Radio Mobile Ad-hoc Networks Jinbei Zhang, Yixuan Li, Zhuotao Liu, Fan Wu, Feng Yang, Xinbing Wang Det of Electronic Engineering Det of Comuter Science and Engineering

More information

Business schools are the academic setting where. The current crisis has highlighted the need to redefine the role of senior managers in organizations.

Business schools are the academic setting where. The current crisis has highlighted the need to redefine the role of senior managers in organizations. c r o s os r oi a d s REDISCOVERING THE ROLE OF BUSINESS SCHOOLS The current crisis has highighted the need to redefine the roe of senior managers in organizations. JORDI CANALS Professor and Dean, IESE

More information

How To Deiver Resuts

How To Deiver Resuts Message We sha make every effort to strengthen the community buiding programme which serves to foster among the peope of Hong Kong a sense of beonging and mutua care. We wi continue to impement the District

More information

Imaginary quadratic orders with given prime factor of class number

Imaginary quadratic orders with given prime factor of class number Imaginary quadratic orders with given rime factor of cass number Aexander Rostovtsev, St. Petersburg State Poytechnic University, [email protected] Abstract Abeian cass grou C(D) of imaginary quadratic

More information

IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 31, NO. 12, DECEMBER 2013 1

IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 31, NO. 12, DECEMBER 2013 1 IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 31, NO. 12, DECEMBER 2013 1 Scaabe Muti-Cass Traffic Management in Data Center Backbone Networks Amitabha Ghosh, Sangtae Ha, Edward Crabbe, and Jennifer

More information

Maintenance activities planning and grouping for complex structure systems

Maintenance activities planning and grouping for complex structure systems Maintenance activities panning and grouping for compex structure systems Hai Canh u, Phuc Do an, Anne Barros, Christophe Berenguer To cite this version: Hai Canh u, Phuc Do an, Anne Barros, Christophe

More information

This paper considers an inventory system with an assembly structure. In addition to uncertain customer

This paper considers an inventory system with an assembly structure. In addition to uncertain customer MANAGEMENT SCIENCE Vo. 51, No. 8, August 2005, pp. 1250 1265 issn 0025-1909 eissn 1526-5501 05 5108 1250 informs doi 10.1287/mnsc.1050.0394 2005 INFORMS Inventory Management for an Assemby System wh Product

More information

Minimizing the Total Weighted Completion Time of Coflows in Datacenter Networks

Minimizing the Total Weighted Completion Time of Coflows in Datacenter Networks Minimizing the Tota Weighted Competion Time of Cofows in Datacenter Networks Zhen Qiu Ciff Stein and Yuan Zhong ABSTRACT Communications in datacenter jobs (such as the shuffe operations in MapReduce appications

More information

Fixed income managers: evolution or revolution

Fixed income managers: evolution or revolution Fixed income managers: evoution or revoution Traditiona approaches to managing fixed interest funds rey on benchmarks that may not represent optima risk and return outcomes. New techniques based on separate

More information

Synopsys RURAL ELECTRICATION PLANNING SOFTWARE (LAPER) Rainer Fronius Marc Gratton Electricité de France Research and Development FRANCE

Synopsys RURAL ELECTRICATION PLANNING SOFTWARE (LAPER) Rainer Fronius Marc Gratton Electricité de France Research and Development FRANCE RURAL ELECTRICATION PLANNING SOFTWARE (LAPER) Rainer Fronius Marc Gratton Electricité de France Research and Develoment FRANCE Synosys There is no doubt left about the benefit of electrication and subsequently

More information

The width of single glazing. The warmth of double glazing.

The width of single glazing. The warmth of double glazing. Therma Insuation CI/SfB (31) Ro5 (M5) September 2012 The width of singe gazing. The warmth of doube gazing. Pikington Spacia Revoutionary vacuum gazing. Image courtesy of Lumen Roofight Ltd. Pikington

More information

A Branch-and-Price Algorithm for Parallel Machine Scheduling with Time Windows and Job Priorities

A Branch-and-Price Algorithm for Parallel Machine Scheduling with Time Windows and Job Priorities A Branch-and-Price Agorithm for Parae Machine Scheduing with Time Windows and Job Priorities Jonathan F. Bard, 1 Siwate Rojanasoonthon 2 1 Graduate Program in Operations Research and Industria Engineering,

More information

Setting Up Your Internet Connection

Setting Up Your Internet Connection 4 CONNECTING TO CHANCES ARE, you aready have Internet access and are using the Web or sending emai. If you downoaded your instaation fies or instaed esigna from the web, you can be sure that you re set

More information

eg Enterprise vs. a Big 4 Monitoring Soution: Comparing Tota Cost of Ownership Restricted Rights Legend The information contained in this document is confidentia and subject to change without notice. No

More information

SELECTING THE SUITABLE ERP SYSTEM: A FUZZY AHP APPROACH. Ufuk Cebeci

SELECTING THE SUITABLE ERP SYSTEM: A FUZZY AHP APPROACH. Ufuk Cebeci SELECTING THE SUITABLE ERP SYSTEM: A FUZZY AHP APPROACH Ufuk Cebeci Department of Industria Engineering, Istanbu Technica University, Macka, Istanbu, Turkey - [email protected] Abstract An Enterprise

More information

3.3 SOFTWARE RISK MANAGEMENT (SRM)

3.3 SOFTWARE RISK MANAGEMENT (SRM) 93 3.3 SOFTWARE RISK MANAGEMENT (SRM) Fig. 3.2 SRM is a process buit in five steps. The steps are: Identify Anayse Pan Track Resove The process is continuous in nature and handed dynamicay throughout ifecyce

More information

Undergraduate Studies in. Education and International Development

Undergraduate Studies in. Education and International Development Undergraduate Studies in Education and Internationa Deveopment Wecome Wecome to the Schoo of Education and Lifeong Learning at Aberystwyth University. Over 100 years ago, Aberystwyth was the first university

More information

Distribution of Income Sources of Recent Retirees: Findings From the New Beneficiary Survey

Distribution of Income Sources of Recent Retirees: Findings From the New Beneficiary Survey Distribution of Income Sources of Recent Retirees: Findings From the New Beneficiary Survey by Linda Drazga Maxfied and Virginia P. Rena* Using data from the New Beneficiary Survey, this artice examines

More information

Early access to FAS payments for members in poor health

Early access to FAS payments for members in poor health Financia Assistance Scheme Eary access to FAS payments for members in poor heath Pension Protection Fund Protecting Peope s Futures The Financia Assistance Scheme is administered by the Pension Protection

More information

Branch-and-Price for Service Network Design with Asset Management Constraints

Branch-and-Price for Service Network Design with Asset Management Constraints Branch-and-Price for Servicee Network Design with Asset Management Constraints Jardar Andersen Roar Grønhaug Mariellee Christiansen Teodor Gabriel Crainic December 2007 CIRRELT-2007-55 Branch-and-Price

More information

A Description of the California Partnership for Long-Term Care Prepared by the California Department of Health Care Services

A Description of the California Partnership for Long-Term Care Prepared by the California Department of Health Care Services 2012 Before You Buy A Description of the Caifornia Partnership for Long-Term Care Prepared by the Caifornia Department of Heath Care Services Page 1 of 13 Ony ong-term care insurance poicies bearing any

More information

The Online Freeze-tag Problem

The Online Freeze-tag Problem The Online Freeze-tag Problem Mikael Hammar, Bengt J. Nilsson, and Mia Persson Atus Technologies AB, IDEON, SE-3 70 Lund, Sweden [email protected] School of Technology and Society, Malmö University,

More information

Insertion and deletion correcting DNA barcodes based on watermarks

Insertion and deletion correcting DNA barcodes based on watermarks Kracht and Schober BMC Bioinformatics (2015) 16:50 DOI 10.1186/s12859-015-0482-7 METHODOLOGY ARTICLE Open Access Insertion and deetion correcting DNA barcodes based on watermarks David Kracht * and Steffen

More information

AA Fixed Rate ISA Savings

AA Fixed Rate ISA Savings AA Fixed Rate ISA Savings For the road ahead The Financia Services Authority is the independent financia services reguator. It requires us to give you this important information to hep you to decide whether

More information

Take me to your leader! Online Optimization of Distributed Storage Configurations

Take me to your leader! Online Optimization of Distributed Storage Configurations Take me to your eader! Onine Optimization of Distributed Storage Configurations Artyom Sharov Aexander Shraer Arif Merchant Murray Stokey [email protected], {shraex, aamerchant, mstokey}@googe.com

More information

Buffer Capacity Allocation: A method to QoS support on MPLS networks**

Buffer Capacity Allocation: A method to QoS support on MPLS networks** Buffer Caacity Allocation: A method to QoS suort on MPLS networks** M. K. Huerta * J. J. Padilla X. Hesselbach ϒ R. Fabregat O. Ravelo Abstract This aer describes an otimized model to suort QoS by mean

More information

Lecture 7 Datalink Ethernet, Home. Datalink Layer Architectures

Lecture 7 Datalink Ethernet, Home. Datalink Layer Architectures Lecture 7 Dataink Ethernet, Home Peter Steenkiste Schoo of Computer Science Department of Eectrica and Computer Engineering Carnegie Meon University 15-441 Networking, Spring 2004 http://www.cs.cmu.edu/~prs/15-441

More information

Learning from evaluations Processes and instruments used by GIZ as a learning organisation and their contribution to interorganisational learning

Learning from evaluations Processes and instruments used by GIZ as a learning organisation and their contribution to interorganisational learning Monitoring and Evauation Unit Learning from evauations Processes and instruments used by GIZ as a earning organisation and their contribution to interorganisationa earning Contents 1.3Learning from evauations

More information

ENFORCING SAFETY PROPERTIES IN WEB APPLICATIONS USING PETRI NETS

ENFORCING SAFETY PROPERTIES IN WEB APPLICATIONS USING PETRI NETS ENFORCING SAFETY PROPERTIES IN WEB APPLICATIONS USING PETRI NETS Liviu Grigore Comuter Science Deartment University of Illinois at Chicago Chicago, IL, 60607 [email protected] Ugo Buy Comuter Science

More information

Qualifications, professional development and probation

Qualifications, professional development and probation UCU Continuing Professiona Deveopment Quaifications, professiona deveopment and probation Initia training and further education teaching quaifications Since September 2007 a newy appointed FE ecturers,

More information

Mean shift-based clustering

Mean shift-based clustering Pattern Recognition (7) www.elsevier.com/locate/r Mean shift-based clustering Kuo-Lung Wu a, Miin-Shen Yang b, a Deartment of Information Management, Kun Shan University of Technology, Yung-Kang, Tainan

More information

Method for Evaluating The Routing Cost in Mpls Network With Regard To Fractal Properties of The Traffic

Method for Evaluating The Routing Cost in Mpls Network With Regard To Fractal Properties of The Traffic Internationa Journa of Soft omputing and Engineering (IJSE) ISSN: 2231-2307 Voume-4 Issue-6 Januar 2015 Method for Evauating The Routing ost in Mps Network With Regard To Fracta Properties of The Traffic

More information

Introduction the pressure for efficiency the Estates opportunity

Introduction the pressure for efficiency the Estates opportunity Heathy Savings? A study of the proportion of NHS Trusts with an in-house Buidings Repair and Maintenance workforce, and a discussion of eary experiences of Suppies efficiency initiatives Management Summary

More information

Vital Steps. A cooperative feasibility study guide. U.S. Department of Agriculture Rural Business-Cooperative Service Service Report 58

Vital Steps. A cooperative feasibility study guide. U.S. Department of Agriculture Rural Business-Cooperative Service Service Report 58 Vita Steps A cooperative feasibiity study guide U.S. Department of Agricuture Rura Business-Cooperative Service Service Report 58 Abstract This guide provides rura residents with information about cooperative

More information

A Supplier Evaluation System for Automotive Industry According To Iso/Ts 16949 Requirements

A Supplier Evaluation System for Automotive Industry According To Iso/Ts 16949 Requirements A Suppier Evauation System for Automotive Industry According To Iso/Ts 16949 Requirements DILEK PINAR ÖZTOP 1, ASLI AKSOY 2,*, NURSEL ÖZTÜRK 2 1 HONDA TR Purchasing Department, 41480, Çayırova - Gebze,

More information

CUSTOM. Putting Your Benefits to Work. COMMUNICATIONS. Employee Communications Benefits Administration Benefits Outsourcing

CUSTOM. Putting Your Benefits to Work. COMMUNICATIONS. Employee Communications Benefits Administration Benefits Outsourcing CUSTOM COMMUNICATIONS Putting Your Benefits to Work. Empoyee Communications Benefits Administration Benefits Outsourcing Recruiting and retaining top taent is a major chaenge facing HR departments today.

More information

Online Media Information

Online Media Information Onine Media Information www.sciencedirect.com The vita daiy source for 16 miion researchers and scientists around the gobe 70 miion pageviews each month Broad advertising product portfoio SciVerse ScienceDirect:

More information

Message. The Trade and Industry Bureau is committed to providing maximum support for Hong Kong s manufacturing and services industries.

Message. The Trade and Industry Bureau is committed to providing maximum support for Hong Kong s manufacturing and services industries. Message The Trade and Industry Bureau is committed to providing maximum support for Hong Kong s manufacturing and services industries. With the weight of our economy shifting towards knowedge-based and

More information

WHITE PAPER BEsT PRAcTIcEs: PusHIng ExcEl BEyond ITs limits WITH InfoRmATIon optimization

WHITE PAPER BEsT PRAcTIcEs: PusHIng ExcEl BEyond ITs limits WITH InfoRmATIon optimization Best Practices: Pushing Exce Beyond Its Limits with Information Optimization WHITE Best Practices: Pushing Exce Beyond Its Limits with Information Optimization Executive Overview Microsoft Exce is the

More information

Jena Research Papers in Business and Economics

Jena Research Papers in Business and Economics Jena Research Paers in Business and Economics A newsvendor model with service and loss constraints Werner Jammernegg und Peter Kischka 21/2008 Jenaer Schriften zur Wirtschaftswissenschaft Working and Discussion

More information

ELECTRONIC FUND TRANSFERS YOUR RIGHTS AND RESPONSIBILITIES

ELECTRONIC FUND TRANSFERS YOUR RIGHTS AND RESPONSIBILITIES About ELECTRONIC FUND TRANSFERS YOUR RIGHTS AND RESPONSIBILITIES The Eectronic Fund Transfers we are capabe of handing for consumers are indicated beow, some of which may not appy your account. Some of

More information

Overview of Health and Safety in China

Overview of Health and Safety in China Overview of Heath and Safety in China Hongyuan Wei 1, Leping Dang 1, and Mark Hoye 2 1 Schoo of Chemica Engineering, Tianjin University, Tianjin 300072, P R China, E-mai: [email protected] 2 AstraZeneca

More information

Example of Credit Card Agreement for Bank of America Visa Signature and World MasterCard accounts

Example of Credit Card Agreement for Bank of America Visa Signature and World MasterCard accounts Exampe of Credit Card Agreement for Bank of America Visa Signature and Word MasterCard accounts PRICING INFORMATION Actua pricing wi vary from one cardhoder to another Annua Percentage Rates for Purchases

More information

Load Balance vs Energy Efficiency in Traffic Engineering: A Game Theoretical Perspective

Load Balance vs Energy Efficiency in Traffic Engineering: A Game Theoretical Perspective Load Baance vs Energy Efficiency in Traffic Engineering: A Game Theoretica Perspective Yangming Zhao, Sheng Wang, Shizhong Xu and Xiong Wang Schoo of Communication and Information Engineering University

More information

NCH Software MoneyLine

NCH Software MoneyLine NCH Software MoneyLine This user guide has been created for use with MoneyLine Version 2.xx NCH Software Technica Support If you have difficuties using MoneyLine pease read the appicabe topic before requesting

More information

HEALTH PROFESSIONS PATHWAYS

HEALTH PROFESSIONS PATHWAYS T heoffic eofcommuni t yco egeres ea r c ha ndl ea der s hi p Co egeofe duc a t i ona ti i noi s The Heath Professions Pathways (H2P) Consortium is a nationa consortium comprised of nine coeges in five states

More information

The Use of Cooling-Factor Curves for Coordinating Fuses and Reclosers

The Use of Cooling-Factor Curves for Coordinating Fuses and Reclosers he Use of ooing-factor urves for oordinating Fuses and Recosers arey J. ook Senior Member, IEEE S& Eectric ompany hicago, Iinois bstract his paper describes how to precisey coordinate distribution feeder

More information

GreenTE: Power-Aware Traffic Engineering

GreenTE: Power-Aware Traffic Engineering GreenTE: Power-Aware Traffic Engineering Mingui Zhang [email protected] Cheng Yi [email protected] Bin Liu [email protected] Beichuan Zhang [email protected] Abstract Current network infrastructures

More information

3.5 Pendulum period. 2009-02-10 19:40:05 UTC / rev 4d4a39156f1e. g = 4π2 l T 2. g = 4π2 x1 m 4 s 2 = π 2 m s 2. 3.5 Pendulum period 68

3.5 Pendulum period. 2009-02-10 19:40:05 UTC / rev 4d4a39156f1e. g = 4π2 l T 2. g = 4π2 x1 m 4 s 2 = π 2 m s 2. 3.5 Pendulum period 68 68 68 3.5 Penduum period 68 3.5 Penduum period Is it coincidence that g, in units of meters per second squared, is 9.8, very cose to 2 9.87? Their proximity suggests a connection. Indeed, they are connected

More information

DAY-AHEAD ELECTRICITY PRICE FORECASTING BASED ON TIME SERIES MODELS: A COMPARISON

DAY-AHEAD ELECTRICITY PRICE FORECASTING BASED ON TIME SERIES MODELS: A COMPARISON DAY-AHEAD ELECTRICITY PRICE FORECASTING BASED ON TIME SERIES MODELS: A COMPARISON Rosario Esínola, Javier Contreras, Francisco J. Nogales and Antonio J. Conejo E.T.S. de Ingenieros Industriales, Universidad

More information

The Whys of the LOIS: Credit Risk and Refinancing Rate Volatility

The Whys of the LOIS: Credit Risk and Refinancing Rate Volatility The Whys of the LOIS: Credit Risk and Refinancing Rate Voatiity Stéphane Crépey 1, and Raphaë Douady 2 1 Laboratoire Anayse et Probabiités Université d Évry Va d Essonne 9137 Évry, France 2 Centre d économie

More information