Dynamic Control of Data Streaming and Processing in a Virtualized Environment

Size: px
Start display at page:

Download "Dynamic Control of Data Streaming and Processing in a Virtualized Environment"

Transcription

1 > REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 1 Dynamc Control of Data Streamng and Proceng n a Vrtualzed Envronment Junwe Cao, Senor Member, IEEE, Wen Zhang, Member, IEEE, and We Tan Abtract Performance of data treamng applcaton co-determned by both networkng and computng reource, and therefore they hould be co-cheduled and co-allocated n an ntegrated and coordnated way. Dynamc control of reource chedulng and allocaton requred, becaue unlateral redundancy n ether networkng or computng reource may reult n the overprovon of t and the other may become a bottleneck. To avod reource hortage a well a overprovon, n th paper, a vrtualzed platform utlzed to mplement data treamng and proceng. In th platform, fuzzy logc controller are degned to allocate CPU reource; teratve bandwdth allocaton appled and proceng- and torage-aware to guarantee on-demand data provonng. Expermental reult how that our approach lead to hgher applcaton performance a well a hgher reource utlzaton, compared wth other reource chedulng and allocaton method. Index Term vrtualzaton, data treamng, reource chedulng and allocaton, fuzzy logc control. Note to Practtoner Data treamng ha become an mportant paradgm n many bune and centfc applcaton, uch a fraud detecton n bankng ndutry and gravtatonal-wave obervaton n atronomy. In th paper, we apply vrtualzaton to provde better upport for thee applcaton, leveragng t advantage n reource on-demand allocaton. A novel dynamc control method propoed o that CPU and bandwdth can be co-cheduled and co-allocated nce for data treamng applcaton thee reource are tghtly coupled from the performance perpectve. In th control method, fuzzy logc control appled for CPU allocaton and an teratve algorthm adopted for proceng-, congeton- and torage-aware bandwdth allocaton. We ue fuzzy control becaue t doe not rely on the mathematcal modelng of an object and can mplement human expert heurtc knowledge va IF-THEN rule. Therefore t derable n our data treamng cenaro, due Manucrpt receved July 10, 2011; reved Oct 16, Th work upported by Natonal Scence Foundaton of Chna (grant No ) and Mntry of Scence and Technology of Chna under Natonal 973 Bac Reearch Program (grant No. 2011CB and No. 2011CB302805) and Natonal 863 hgh-tech Program (grant No. 2011AA040501). Junwe Cao wth Reearch Inttute of Informaton Technology and Tnghua Natonal Laboratory for Informaton Scence and Technology, Tnghua Unverty, Bejng , Chna (correpondng author, phone: ; fax: ; e-mal: jcao@tnghua.edu.cn). Wen Zhang wa wth Natonal CIMS Engneerng Reearch Center, Tnghua Unverty, Bejng , Chna. He now wth Chongqng Mltary Delegate Bureau, General Armament Department of PLA, Chongqng , Chna. We Tan wth IBM T. J. Waton Reearch Center, Hawthorne, New York, USA (e-mal: wtan@u.bm.com). to the varable couplng and heavly nonlnear nature of the ytem. I. INTRODUCTION ata treamng and proceng ha become more mportant Dn many bune and centfc applcaton. Such applcaton requre effcent tranmon of data from/to dtrbuted ource. It often not feable to tore the entre data before ubequent proceng becaue of the lmted torage and hgh volume of data to be proceed. Mot applcaton run contnuouly and requre effcent ue of computatonal reource to carry out proceng n a tmely manner [1]. An example the LIGO (Laer Interferometer Gravtatonal-wave Obervatory) [2] whch am at the detecton of gravtatonal wave emtted from pace ource. LIGO data analy tream terabyte of data per day from obervatore for real-tme proceng ung centfc workflow [3]. Th a typcal cenaro n many centfc applcaton where data proceng contnuouly conducted over remote tream a f data were alway avalable from local torage. In our prevou work [6], we have already demontrated that, n data treamng applcaton, unlateral redundancy of ether CPU or bandwdth doe not necearly lead to hgh throughput; on the other hand, hortage of ether may alo negatvely mpact the throughput. Th reveal the tght couplng between CPU and bandwdth from the performance perpectve. Therefore t requred to allocate computng and networkng reource n order to reach a balance between hgh throughput and hgh reource utlzaton. Vrtualzaton technology whch can dynamcally provon vrtual machne (VM) [5], a natural choce for reource management [4] n th cenaro. The contrbuton of th work twofold. Frt a new co-chedulng framework propoed on a vrtualzed platform, o that CPU and bandwdth can be co-allocated for LIGO tream proceng. Such co-chedulng addree the mpact of the nterplay between CPU and bandwdth on the ytem performance. Second, a fuzzy-logc baed, cloed-loop feedback control [7] method deved for the aforementoned co-chedulng framework. Ung th method we control VM CPU allocaton by confgurng VM dynamcally accordng to reource utlzaton, and teratvely allocate bandwdth by cloely watchng proceng and torage tatu. The ret of th paper organzed a follow: Secton 2 formulate the data treamng and proceng problem, and

2 > REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 2 Secton 3 and 4, fuzzy allocaton of CPU reource and teratve bandwdth allocaton are dcued, repectvely. Expermental reult are llutrated n Secton 5 to how performance evaluaton of our approach ung a gravtatonal wave data analy applcaton. Secton 6 dcue related work and Secton 7 conclude the paper. II. DATA STREAMING AND PROCESSING A. Data Streamng Applcaton Many extng centfc applcaton requre data treamng and proceng n a real tme manner. Data ource can be large-cale mulator or obervatore, wth megabyte of data generated per econd and terabyte of data aggregated per day. Data are uually treamed to remote proceng node for varou analye. For thee proceng node, t not feable to tore all data nce new data contantly arrve and conume local tore pace. Therefore, after data are proceed and become obolete, they need to be removed for newly arrval data. Th typcal cenaro reult n a tghtly coupled relatonhp among computng, torage and networkng reource. For example, a LIGO gravtatonal wave data analy applcaton read n two data tream from two remote LIGO obervatore (one n Wahngton State and the other n Louana State) and calculate correlaton coeffcent that can be ued to characterze mlarty of two data curve. If two gnal from two obervatore occur multaneouly wth mlar curve, t would be lkely that a gravtatonal wave canddate detected. LIGO data are archved n pecally formatted bnary fle,.e., Gravtatonal Wave Framefle (gwf). Each fle compoed wth 16 or 256 of data from multple channel of an obervatory. A LIGO data analy applcaton uually nvolve multple data tream (ere of mall data fle) from multple obervatore. LIGO data are archved n LIGO data grd [43] node whch cannot provde enough computng reource for drectly local data analy. LIGO tre to beneft from open computng reource uch a the Open Scence Grd (OSG) [44]. Whle abundant computng reource are avalable n OSG, no LIGO data are avalable on OSG node. It become crucal to tream LIGO data from LIGO data grd node to OSG reource for large-cale proceng. Data treamng and proceng become eental for LIGO data analy [38][39]. LIGO data analy applcaton are developed ung dfferent operatng and programmng envronment. In order to acheve fner-graned reource chedulng and hgher utlzaton of computng reource, multple applcaton have to further hare one phycal machne, and thu vrtualzaton become the enablng technology. A vrtualzed envronment can hot multple uch data treamng applcaton. One ngle vrtual machne (VM) provde a predctable and controllable run-tme envronment for each applcaton. All computng, torage and networkng reource on a vrtualzed platform can be hared among multple VM (and ultmately multple applcaton), a llutrated n Fg. 1. Whle our prevou work wa focued on reource harng among multple phycal machne [1], n th work, a vrtualzed envronment deployed, whch brng new challenge on dynamc control. Fg. 1. An llutraton of reource harng n a vrtualzed envronment among multple data treamng applcaton. The local torage play a key role correlatng networkng and computng reource. If no data avalable, computng reource wll be dle. If the allocated local torage full for an applcaton, data treamng cannot be carred out and networkng reource cannot be utlzed. At any tme t, for a data treamng applcaton, f the amount of data n local torage, denoted a Q (t), hgher than a certan level (e.g., a block a explaned later), data proceng trggered. Q (t) co-determned by both data provonng and proceng nce new data wll be treamed to local torage whle proceed data wll be cleaned up afterward. The amount of output data (e.g. tattcal value) uually mnor and gnored when we calculate local torage. The amount of data n torage vare over tme and can be decrbed ung the followng dfferental equaton: Q t tranpeed t d t, where ( t) Q Reource n a Vrtualzed Envronment Bandwdth Dk CPU Vrtual reource for app 1 Vrtual reource for app 2 Vrtual reource for app Vrtual reource for app n Q ( 0 ) 0 (1), tranpeed (t) and d (t) tand for the dervatve of Q (t), agned tranferrng bandwdth and proceng peed for data tream. If there are data avalable n the local torage, an ndcator, denoted a Ready for the applcaton, et to be 1, otherwe Ready 0. So d (t) can be decrbed a: 0, Ready 0 d () t > 0, Ready 1 B. Performance Metrc For data treamng applcaton, data throughput the mot mportant performance metrc. Meanwhle reource utlzaton hould be alo condered. Real Proceng Speed (RPS): the actual data proceng peed gven by d (t) Theoretc Proceng Speed (TPS): the data proceng peed the allocated CPU reource can generate f there were alway uffcent data provoned, denoted a

3 > REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 3 procpeed (t,c (t)), where C (t) tand for the allocated CPU reource for applcaton at tme t. Relatonhp between procpeed (t,c (t)) and C (t) mut be determned wth ytem dentfcaton and t obvou that procpeed (t,c (t)) a non-decreang functon of C (t), where C (t) manly refer to a proporton of CPU cycle a explaned later. Real Throughput (RTP): gven a data provonng cheme, the actual amount of data proceed n a gven perod of tme. Theoretc Throughput (TTP): the amount of data proceed n a gven perod of tme f there were alway enough data provonng. Schedulng CPU, torage and bandwdth reource carred out perodcally to deal wth dynamc nature of reource and applcaton, and each perod referred to a a chedulng perod. Suppoe the length of a chedulng perod M, and for the h th chedulng perod, the followng formula are traghtforward: d 0, Re ady procpeed t, C () t () t d t, C () t From (1): TTP, h ( 1) hm M 0 { ( ), Re ady 1 procpeed ( t, C () t ) ( h 1) RTP hm, h d ( t, C () t ) dt ( h 1) M hm. ( () () ( h 1) M ) RTP h, tranpeed t Q t dt hm tranpeed t dt+ Q h M Q hm t h M dt () (( 1) ) Defne utlzaton of computng reource (UC n hort) a RTP, h UC, h TTP (2), h.e., UC hm tranpeed () t dt+ Q ( h 1) M Q hm t ( h 1 ) M, h hm (3) procpeed () t dt t ( h 1 ) M, denotng to what extent the allocated compute reource utlzed. RTP,h can be defned n another form a: RTP, h procpeed () t dt (4) Ω, h where Ω,h tand for the tme fragment when proceng gong on and then utlzaton can be redefned n another way a: Ω h UC, h M (5) Note that (5) mple that TPS,h a contant n a chedulng perod wth gven CPU reource. UC can be defned alo a the rato of RPS to TPS. The problem to allocate proper amount of CPU reource to generate RPS approachng TPS a much a poble gven the data upply cheme. It obvou that redundant CPU reource wll make a TPS much larger than RPS, whch mple underutlzaton of computng reource. If avalable bandwdth lmted, RPS wll be zero at mot tme wth redundant CPU cycle for lack of data to proce. Th dependency between data provon and proceng make t neceary to allocate compute reource on demand o a to make RPS a cloe to TPS a poble. III. CPU ALLOCATION WITH FUZZY CONTROL CPU allocaton mplemented ung a fuzzy control approach on top of vrtualzaton technology, where the vrtualzaton provde an olated run-tme envronment and the fuzzy control addree approprate reource confguraton n a vrtualzed envronment. A. Vrtualzaton wth Xen Recent progre on vrtualzaton technology make t poble for reource olaton and performance guarantee for each data treamng applcaton. Vrtualzaton provde a layer of abtracton n dtrbuted computng envronment, and eparate phycal hardware wth operatng ytem, o a to mprove reource utlzaton and flexblty. Xen [36], an open ource hypervor, ue to buld the vrtualzed envronment. Wth Xen, confguraton of VM can be dynamcally adjuted to optmze performance. The CPU of a VM called vrtual CPU, often abbrevated a VCPU. The quota of phycal CPU cycle a VCPU wll get determned by two parameter,.e., cap and weght. The cap value defne the maxmum percentage of the CPU that can be ued by the VM; when multple VM compete for one CPU, ther weght value defne ther proporton n gettng the hared CPU. For example, a VCPU wth a weght of 128 can obtan twce a many CPU cycle a one whoe weght 64, whle 50 a a cap value ndcate that the VCPU wll obtan 50% of a phycal CPU cycle. In th work, cap adjuted dynamcally accordng to the meaured utlzaton and pre-defned fuzzy rule a decrbed below. B. Fuzzy Control A fuzzy control ytem [37] baed on fuzzy logc related wth fuzzy concept that cannot be expreed a true or fale but rather a partally true. A fuzzy logc controller (FLC) depcted n Fg. 2; t cont of an nput tage, a proceng tage, and an output tage. Some bac concept are gven below to help contruct an elementary undertandng of fuzzy logc controller and ther mechanm. Unvere of dcoure the doman of an nput (output) to (from) the FLC. Input and output mut be mapped to the unvere of dcoure by quantzaton factor (Ke and Kec n Fg. 2) and calng factor (Ku n Fg. 2), repectvely, whch help to mgrate the fuzzy control logc to dfferent problem wthout any modfcaton.

4 > REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 4 Fg. 2. A fuzzy logc controller. The nput varable n a fuzzy control ytem are n general mapped nto fuzzy et, where an nput varable may be mapped nto everal fuzzy et wth correpondng truth value determned by the memberhp functon. Th proce called fuzzfcaton. All the rule that apply are nvoked, ung the memberhp functon and truth value obtaned from the nput, to determne the reult of the rule. Th reult n turn wll be mapped nto a memberhp functon and truth value controllng the output varable. Thee reult are combned to gve a pecfc anwer by a procedure known a defuzzfcaton. Lngutc varable decrbe the nput and output() of a fuzzy controller. Thee lngutc varable are a natural way reemblng human thought to handle uncertante. Lngutc varable nvolved n th work nclude the nput of the fuzzy controller, UC and UC and the output of the fuzzy controller, a proportonal factor (PF). Lngutc value are ued to decrbe charactertc of the lngutc varable. Very low, low, medum, hgh and very hgh are the lngutc value for UC, whle thoe for UC and PF are NB, NM, NS, ZE, PS, PM and PB, where N, P, B, M, S and ZE are abbrevaton of negatve, potve, bg, medum, mall and zero, repectvely, and the combnaton of them jut take on a degree of truth. Dfferent from clacal mathematc, n fuzzy world, th repreented a a contnuou value between 0 to 1, and 0.5 ndcate we are halfway certan. The mappng from a numerc value to a degree of truth for a lngutc value done by the memberhp functon. Lngutc rule form a et of IF preme THEN conequent rule to map the nput to output() of a fuzzy controller,.e., to gude the fuzzy controller acton. Thee rule are defned n term of lngutc varable, dfferent from the numercal nput-or-output of the clacal controller. A lngutc rule for example : IF UC hgh AND UC NB THEN PF NB. Rule-bae hold a et of IF-THEN rule a a part of the controller, dctatng how to acheve PF accordng to the fuzzfed lngutc value of UC and UC. Memberhp functon quantfy the certanty an UC and UC value to be aocated wth a certan lngutc value. Except for the memberhp of lngutc value very low for UC, we ue ymmetrc trangle of an equal bae and 50% overlap wth adjacent MF. Unlke tradtonal et theory, n fuzzy et theory underlyng fuzzy control theory, et memberhp not bnary but contnuou to deal wth uncertante. Thu, a fuzzy nput or output may belong to more than one et maxmum two adjacent et n our MF wth dfferent certanty value. Inference mechanm n Fg. 2 determne whch rule wll be appled at the k th amplng pont, baed on the fuzzfed UC and UC. To compute the certanty value of the preme n the correpondng IF preme THEN conequent rule(), we take the mnmum between the certanty value of UC and UC, nce the conequent cannot be more certan than the preme. Fuzzfcaton to tranform prece value of nput nto fuzzy et wth correpondng memberhp functon, whch ndpenable for fuzzy nference. Output of fuzzy nference mut be tranformed nto a clear value by defuzzfcaton. C. Lngutc Varable and Fuzzy Rule A for the FLC propoed n th paper, the nput are the oberved reource utlzaton UC a defned n (2) or (5) and the dervatve of t, UC. Although t ha two nput, eentally t a ngle nput controller nce UC can be derved from UC, UCUC(k)-UC(k-1). The output of the fuzzy controller for applcaton n the h th chedulng perod a proportonal factor, denoted a PF,h. At the h th chedulng perod, gven nput UC and UC, uppoe the relevant fuzzy et of output form a et denoted a M,h wth memberhp denoted a M,h (u), where u U,h and U,h the unvere of dcoure, then the output can be calculated a PF M u udu M u du (6) h, h, h, Uh, Uh, Suppoe the ntal cap of each applcaton are C,0, n the h th chedulng perod, the cap wll be ( PF 1) CapScale, 1 C, h C, h 1 +, h 1 h (7) where CapScale the varyng cale of the allocated cap. PF,h adjuted every chedulng perod, o t adaptve to the varyng tuaton. Relatonhp between the allocated cap and procpeed a lnear model, a decrbed later n Secton V. p q procpeed() h a l procpeed( h l) + bmcap( h m) (8) l 1 m 1 Trangular memberhp functon are adopted for nput UC and UC and output PF, a hown n Fg. 3. Table I provde a ummary of fuzzy rule. A low utlzaton mple that the allocated CPU quota hould be decreaed to releae redundant compute reource wthout reducng the ultmate throughput; meanwhle extremely hgh utlzaton ndcate that more CPU reource are requred to ncreae proceng effcency. When the utlzaton very low, low or medum, the generated PF hould be le than 1 whle the very hgh utlzaton requre a PF lager than 1. To avod ocllaton, UC hould alo be pad attenton. When the utlzaton far away from the ettled pont (80% here), the adjutment can be bg. Then the PF n the 1 t, 2 nd and 3 rd column n Table I wll be NB (negatvely bg) whle n the lat column t PB (potvely bg). When the utlzaton fall to the hgh area, more careful adjutment requred, a hown n the 4 th column n Table I. For example, when UC NS, PF alo NS; and when UC ZE, PF alo ZE, whch mean that no adjutment requred o a to keep a table tatu. From Table 1, t can be een that thee fuzzy rule are mple but robut. It can guarantee a rapd convergence to the ettled

5 > REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 5 pont wthout table tate error, whch a requred charactertc for control ytem a hown n expermental reult ncluded n Secton V very low low medum hgh very hgh Input varable UC A FLC receve UC and UC and output the cap of CPU for each applcaton and then procpeed determned ung (8). An teratve bandwdth allocaton (IBA) mplemented a decrbed n Secton IV to decde tranpeed. UC at the next chedulng perod obtaned wth (2) or (5). In uch a way, the control ytem work and dynamc reource allocaton mplemented for vrtualzaton. Fg. 4. Dynamc control of CPU and bandwdth reource co-chedulng and co-allocaton. Fg. 3. Trangular memberhp functon of nput UC and UC and output PF. Lngutc value of UC nclude very-low, low, medum, hgh and very-hgh, ndcatng CPU utlzaton. Both nput UC and output PF adopt trangular memberhp functon wth lngutc varable of NB, NM, NS, ZE, PS, PM and PB. The unvere of dcoure of UC fall to the cope of -0.4 to 0.4, whch baed on our emprcal obervaton. It alo the cae for PF where the unvere of cope et to 0.6 to 1.4. UC PF NB NM NS ZE PS PM PB TABLE I. FUZZY RULES UC very low low medum hgh very hgh NB NB NB NB NM NS ZE PS PM PB D. Reource Co-chedulng and Co-allocaton Dynamc control of reource co-chedulng and co-allocaton for data treamng applcaton llutrated n Fg. 4. CPU and bandwdth reource are co-allocated by the actuator ACT. The tranfer functon G from allocated CPU and bandwdth reource to utlzaton UC not avalable, for the two nput are tghtly coupled wth each other. Fortunately, th tranfer functon not ndpenable for our fuzzy allocaton cheme. PB IV. ITERATIVE BANDWIDTH ALLOCATION Snce data are treamed to local torage through network, multple data tream need to hare the total bandwdth of the vrtualzed envronment, denoted a I. The ndvdual data tream, called eon, denoted a, form a et S. Each eon wll be agned wth a bandwdth x (.e., tranpeed n Secton II), where x X, X [b, B ] and b >0, B <. b tand for the leat bandwdth requred for eon, whle B the hghet bandwdth of the connecton from the correpondng data ource to eon. Seon ha a utlty functon U (x ) that aumed to be concave, contnuou, bounded and ncreang n the nterval [b, B ]. We try to maxmze the um of the utlte of all the eon, mantanng farne among them. The problem can be decrbed a follow. max U ( x ) S S.t. x I x X Varable U and L are the pre-defned upper and lower bound of data volume n torage for eon to control the paue and reumng of data tranfer,.e., when the data volume n torage of reache the upper bound, data tranfer halted and when th volume reache the lower bound, tranfer reumed. By th mean data tranfer may be ntermttent rather than contnuou, o a to be torage aware, to avod data overflow whle guaranteeng data provonng. Accordng to the amount of data n torage, there are two poble tranfer tate for each at any tme,.e., actve and nactve. All actve eon form a et, called S A, and t obvou that th et varyng becaue tranfer tate of eon are changng. At every amplng tme k, tatu of tranfer can be determned by data amount n torage and the prevou tatu:

6 > REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 6 tatu, k 1, amount, k 1 < U, tatu, k 1 1 0, amount, k 1 U 0, amount, k 1 > L, tatu, k 1 0 1, amount, k 1 L, where tatu,k and amount,k tand for tatu of tranfer and data amount n torage for eon at the k th amplng tme, repectvely. The ntal condton tatu,0 1 and amount,0 0. We only allocate bandwdth for actve tranmon, o the bandwdth contrant become: x I. S A An teratve optmzaton algorthm propoed a follow. Whle S A ( k + 1) x otherwe, ( [ x k + k ) ( k ) α U ' k x ] f x X S A β ( k ) k ρ [ x ] f x x k ( k+ 1 ) X 0, S S A A > I ρi Here, x (k) the bandwdth for eon S at the k th amplng tme. {α k } and {β k } are two potve equence, β k (0, 1). [ ] X denote a projecton on the et X : y mn B, max b y '( ) U the ub-gradent of [ ] ( ) X, ( k ) () U x U and ' ( k ) U x S A U () x And ρ the o-called afety coeffcent to avod bandwdth exce, where ρ (0, 1). Th allocaton algorthm alo proceng-aware, nce tatu,k and amount,k are affected by proceng,.e., S A aocated wth data proceng. On the other hand, the allocated bandwdth wll alo change UC to trgger the FLC to adjut allocaton of CPU reource. In th way, bandwdth and CPU reource are co-cheduled and co-allocated. Parameter n th teratve bandwdth allocaton cenaro, uch a α k, β k and ρ can be adjuted accordng to dfferent allocaton prncple, uch a relatve farne and the-mot-needed-the-frt. The mot appled utlty functon logarthmc: U ( x ) w ln ( 1+ x ), where w the weght of eon x and a larger w mple a bgger quota n the total avalable bandwdth. Due to the couplng of data provonng and proceng, the relatonhp between proceng effcency and allocated CPU or bandwdth heavly non-lnear and hard, f not mpoble, to be expreed wth a cloed-form formula. It dffcult to apply tradtonal feedback control n abence of a prece model, and that why n Secton III we propoed to ue fuzzy control a an alternatve, due to t model-free nature. ( k ) V. PERFORMANCE EVALUATION A. Experment Confguraton VM are et up on a HP DL580G5 erver wth 4 Intel CPU contanng 16 Xeon E7310 core and 8 GB memore for LIGO data treamng applcaton. Data tem are treamed to the VM from remote data ource. LIGO data tream are compoed wth numerou mall data fle, each contanng obervatonal data acqured n 16 econd. Here 1,188 par of LIGO data fle from two obervatore wth the total amount of 4,354 MB are ued n the experment a ample data. In the followng experment, thee data are utlzed perodcally when needed. To reveal the mathematcal relatonhp between the proceng peed and the allocated computng reource (manly the quota of CPU), ytem dentfcaton carred out. The makepan durng whch all the data par are proceed are hown n Fg. 5, from whch t eem that allocated memory ha lttle mpact on the makepan for th applcaton. Makepan () Fg. 5. Makepan wth dfferent cap and memory ze. Three VM are etup wth 512 MB, 256 MB and 128 MB of memory, repectvely. The allocated cap for each VM range from 5% to 100%. The proceng peed hown n Fg. 6 wth the old lne. Polynomal curve fttng appled to generate a mathematcal functon from the cap to proceng peed, ung the Leat Square Method (LSM). Proceng Speed (M/) Meaured Ftted Cap (%) Fg. 6. Proceng peed wth dfferent cap. Memory ze for each VM et to 128 MB. Fg. 5 and 6 nfer that once the allocated cap exceed 50%, the ame ncreae n CPU Cap wll not gan o obvou beneft a t doe n the range of 5%-50%. In our experment, applcaton can be dvded nto two categore: lght and heavy one. In lght applcaton, data

7 > REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 7 tem are proceed wth mall amount of computaton whle heavy applcaton perform more complex data proceng and relatvely larger amount of computatonal reource are requred. Heavy and lght group of applcaton are ued a benchmark, each group wth 5 applcaton. Total avalable bandwdth I, et to 5, 10 and 15 Mbp, hared among all applcaton. Three value of CapScale, 3, 5 and 8, are evaluated. Our allocaton algorthm evaluated for 100 chedulng perod. Each perod lat for 100 econd n thee experment. Performance metrc nclude the output of the FLC (PF), the utlzaton of each allocated computatonal reource (a defne n (1), UC), the allocated cap for each applcaton and reource uage (US) of CPU and bandwdth. Note that the uage of CPU mean the um of allocated cap for each applcaton, whch dfferent from the utlzaton defned n (1). B. Reource Allocaton and Utlzaton Fg. 7 and 8 llutrate the proportonal factor (PF), allocated reource (UC), cap value (Cap) and reource uage (US) value of heavy and lght applcaton, repectvely. A hown n Fg. 7 and 8, lght and heavy applcaton get approprate amount of CPU reource ung our approach, where the total bandwdth 5 Mbp. In thee cae, requred CPU reource of each applcaton far le than a whole phycal CPU, becaue the allocated cap of them are under 20% or even 10%. The total CPU utlzaton alo far le than 100% app1 app2 app3 app4 app5 (c) (d) Fg. 7. Performance of heavy tak. (a) Proportonal factor; (b) Utlzaton; (c) Allocated cap; (d) Reource uage. Intal cap for each applcaton are 10%. All the allocaton cheme are converged to a teady tate. No teady tate error ext n each allocaton cheme. Alo n preence of udden change of avalable reource, they can make a rapd repone, a hown n Fg. 8. PF Schedulng Perod (a) (a) (b) Fg. 7. (to be contnued) (b) Fg. 8. (to be contnued)

8 > REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 8 (c) (c) (d) Fg. 8. Performance of lght tak. (a) proportonal factor; (b) utlzaton; (c) allocated cap; (d) reource uage. C. Robutne and Adaptablty Adaptablty of the allocaton algorthm alo teted. Total avalable bandwdth jump to 8 and 10 at the 30 th and 60 th chedulng perod, repectvely. Performance metrc are ued the ame a thoe n Fg 7-8, but here we only how reult wth heavy tak. (a) (d) Fg. 9. Reponvene to bandwdth change wth heavy tak. (a) proportonal factor; (b) utlzaton; (c) allocated cap; (d) reource uage. A hown n Fg. 9, durng the frt 30 expermental perod, reult are the ame a thoe ncluded n Fg. 7. After 30 perod when avalable bandwdth jump, our control method can react to bandwdth change very fat. From the perpectve of control theory, th varance n bandwdth can be condered a a dturbance, and Fg. 9 how that the robutne, or ant-dturbance capablty of our approach, atfactory. D. Parametrc Convergence and Stablty A a routne n control ytem degn, tablty analy ndpenable becaue only table ytem can be appled. We fnd that ytem tablty entve to ome parameter, e.g. CapScale n (7). A larger CapScale that reult n a rapd convergence to a teady tate may alo lead to performance ocllaton and ntablty. For example, when CapScale et to 8, lght applcaton wll not converge to a table value. In th cae, the control ytem cannot work n a table tate, a hown n Fg. 10. The reaon that even the mallet control cale already exceed requred control power, o the ytem are adjuted ether above or below and never reach the table tatu. (b) Fg. 9. (to be contnued)

9 > REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 9 (a) (b) UC Cap (b) (c) Fg. 10. Performance of lght tak wth a large CapScale. (a) proportonal factor; (b) utlzaton; (c) allocated cap. Heavy tak are not a entve to a large CapScale a lght one nce ther tolerance to the mallet CapScale hgher. In Fg. 11, CapScale alo et to 8. Whle ome ocllaton occur, the magntude very mall. In our applcaton we et CapScale from 3 to 5 baed on our experence. PF app1 app2 app3 app4 app Schedulng Perod (a) Fg. 11. (to be contnued) (c) Fg. 11. Performance of heavy tak wth a large CapScale. (a) proportonal factor; (b) utlzaton; (c) allocated cap. E. Performance Comparon Several other reource allocaton cheme are developed n comparon, a hown n Table II: teratve and even tand for the way to allocate bandwdth among applcaton, n an teratve way a decrbed n Secton IV or jut to dvde the total bandwdth evenly; dynamc tand for CPU allocaton manner decrbed n Secton III, whle fxed mean that allocated CPU reource for each applcaton are contant. Obvouly, our approach belong to Cae 1. TABLE II. ALGORITHM SETTINGS Bandwdth CPU Cae 1 Iteratve dynamc Cae 2 Iteratve Fxed Cae 3 Even dynamc Cae 4 Even Fxed Some reult are provded n Table III where performance metrc from top to bottom are fnal throughput (.e., the volume of data proceed durng the experment, and TP n hort), CPU uage and bandwdth (BD n hort) uage n percentage. Character of H and L are abbrevaton of heavy and lght, ndcatng the type of applcaton. Stll experment are carred out for 100 perod and each perod lat for 100 econd. Therefore, each cenaro nvolve an upper lmt of 50, 100 and 150Gb of data tranfer, repectvely, correpondng to three total bandwdth ettng,.e., 5, 10, and 15Mb. Our algorthm preval n all the cenaro n Table III. For example, ung fewer reource, our algorthm obtan a hgher throughput. In the extreme cae,.e., Cae 4 where the bandwdth evenly allocated and CPU reource are fxed, the performance not atfactory depte the fact that ther reource conumpton ubtantal. The performance of Cae

10 > REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 10 2 and 3 better than Cae 4, but tll not a good a our. Th reult quanttatvely valdate our earler aumpton that unlateral adjutment of bandwdth or CPU reource not uffcent enough to reach the goal of hgh throughput and hgh utlzaton of reource multaneouly. Furthermore t jutfe our methodology to co-chedule bandwdth and CPU reource n an ntegrated and coordnated way. Index Fnal TP (Mb) CPU Uag e (%) BD Uag e (%) TABLE III. PERFORMANCE COMPARISON Cae Total Bandwdth (Mbp) H L H L H L VI. RELATED WORK Stream proceng [8] ha been one of the major focue of databae reearch n recent year, and ome technque, e.g. Aurora [9] and TelegraphCQ [10], have been developed to deal wth contnuou quere on data tream. Thee ytem often deal wth lght-weght data,.e., the volume of each tem n the ncomng tream mall and uually n-memory proceng needed to yeld an extremely low latency. Our work focue on chedulng data treamng job on vrtualzed reource. Dtrbuted computng technque evolve from cluter, grd to cloud computng [11]. Reource management and allocaton ha been a key ue n thee area [12]. In cloud computng wth vrtualzaton technology a the key enabler, vrtual machne [13] or vrtual cluter [14] are bac unt n management, chedulng and optmzaton [15]. Tool ncludng Eucalyptu [16], VMPlant [17] and Uher [18] can erve th management purpoe. Some cheduler are developed to upport data treamng applcaton, e.g. GATES [19] and Streamlne [20], but they manly concern on computng reource allocaton wthout takng bandwdth nto account. Several project uch a EnLIGHTened [21], G-lambda [22] and PHOSPHORUS [23] put emphae on networkng reource. They have the total control over a dedcated optcal network o that a determntc path can be obtaned wth advance reervaton or on demand, whle our work ue a TCP/IP baed hared network, uch that bandwdth more of an ue. Streamflow [40] propoe a framework that enhance a tradtonal centfc workflow nfratructure to enable the tream proceng capablty, and reource allocaton not dcued. In [41] a dtrbuted and ppelned dataflow executon ytem propoed. The executon optmzed by explotng parallelm, load balancng and co-locatng. In our approach all the executon n a vrtual envronment and we have a unque requrement to harmonze computaton and bandwdth allocaton. Control theory ha been uccefully appled to optmze performance or qualty of ervce (QoS) for varou computng ytem. An extenve ummary of related work can be found n the frt chapter of th book [7]. Approache uch a proportonal, ntegral, and dervatve (PID) control [24][25], pole placement [26], lnear quadratc regulator (LQR) [26] and adaptve control [27][28] are appled n varou ytem. Mot of them requre a prece model of the controlled object(). Ung thee approache t nevtable to etablh dfferental equaton model that defne the ytem repone to nput. Th rather challengng n ome cae ncludng data treamng applcaton, due to varable couplng and heavly nonlnear property of the ytem. Fortunately, fuzzy control offer an alternatve, provdng a formal technque to repreent, manpulate and mplement human expert heurtc knowledge for controllng an object va IF-THEN rule. Fuzzy control doe not rely on mathematcal modelng of an object and ntead t etablhe a drect nonlnear mappng between nput and output. Th can gnfcantly reduce the dffculty of a control ytem degn n our applcaton cenaro. The frt applcaton of fuzzy control wa ntroduced nto ndutry n 1974 [29]. Fuzzy control [30][31] alo a reearch topc n computng ytem manly on admon control to get a better qualty of ervce. Adaptve fuzzy control appled for utlzaton management of perodc tak [32], where the utlzaton defned a the rato of the etmated executon tme to the tak perod. Fuzzy nference carred out to decde the threhold over whch the QoS of tak hould be degraded or even ncomng tak be rejected. In th work, executon tme etmaton mut be provded, whch not feable for ome applcaton. A recent tudy relevant to our work [33] focued on provdng predctable executon o a to meet the deadlne of tak. Vrtualzaton technology appled to mplement the o-called performance contaner and compute throttlng framework, to realze the controlled tme-harng of hgh performance computng reource. Sytem dentfcaton carred out to etablh the model of controlled object and a proportonal and ntegral (PI) controller ued. Th work ha mlar motvaton wth our, whle we adopt the model-free fuzzy control approach gven the nature of the condered data

11 > REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 11 treamng applcaton. VII. CONCLUSION In th work we provde a new approach for co-chedulng and co-allocaton of vrtualzed reource for data treamng and proceng. Fuzzy logc control of CPU reource and teratve bandwdth allocaton are mplemented for ntegrated and coordnated reource management. Expermental reult how the good performance n reource utlzaton, data throughput and robutne of the approach, n comparon wth everal other method wth le adaptablty to dynamc envronment. In further work, bede CPU and bandwdth, we plan to nvetgate vrtualzed network and elatc torage allocaton. More complex cenaro, uch a workflow [34][42], ervce compoton [35], and economcal provonng [47] wll be condered. In thee cenaro VM can be ued for each tage n a proce and optmzed to acheve a balanced reource allocaton and thu a better overall performance. Vrtualzaton technology ha been appled n the LIGO communty for data analy oftware development [45]. In future vrtualzaton baed dynamc control wll become more crucal for larger-cale runtme LIGO data treamng and proceng, nce the development of next generaton of gravtatonal wave detector [46] wll nvolve much larger amount of data and requre more computaton and network reource. ACKNOWLEDGMENT We thank the anonymou revewer for detaled and contructve comment. Th work carred out n cloe collaboraton wth LIGO (Laer Interferometer Gravtatonal-wave Obervatory) Laboratory at Maachuett Inttute of Technology. Junwe Cao would lke to expre h grattude to Dr. Erk Katavound for h long-term upport on LIGO gravtatonal wave data analy work at Tnghua Unverty. REFERENCES [1] W. Zhang, J. Cao, Y. Zhong, L. Lu, and C. Wu, Grd Reource Management and Schedulng for Data Streamng Applcaton, Computng and Informatc, Vol. 29, pp , [2] A. Abramovc, W. E. Althoue, et. al., LIGO: The Laer Interferometer Gravtatonal-Wave Obervatory, Scence, Vol. 256, No. 5055, pp , [3] D. A. Brown, P. R. Brady, A. Detz, J. Cao, B. Johnon, and J. McNabb, A Cae Study on the Ue of Workflow Technologe for Scentfc Analy: Gravtatonal Wave Data Analy, n I. J. Taylor, D. Gannon, E. Deelman, and M. S. Sheld (Ed.), Workflow for escence: Scentfc Workflow for Grd, Sprnger Verlag, pp , [4] M. Mglard, J. Dongarra, A. Get and V. Sunderam, Dynamc Reconfguraton and Vrtual Machne Management n the Harne Metacomputng Sytem, Computng n Object-Orented Parallel Envronment, Lecture Note n Computer Scence, Vol. 1505, pp , [5] P. Barham, B. Dragovc, K. Fraer, S. Hand, T.L. Harr, A. Ho, R. Neugebauer, I. Pratt and A. Warfeld, Xen and the Art of Vrtualzaton, Proc. ACM Symp. on Operatng Sytem Prncple, [6] W. Zhang, J. Cao, Y. Zhong, L. Lu and C. Wu, An Integrated Reource Management and Schedulng Sytem for Grd Data Streamng Applcaton, Proc. 9 th IEEE/ACM Int. Conf. on Grd Computng, pp , Tukuba, Japan. [7] J. L. Hellerten, Y. Dao, S. Parekh, and D. M. Tlbury, Feedback Control of Computng Sytem, Wley-IEEE Pre, Augut [8] L. Golab, and M. T. Ozu, Iue n Data Stream Management, SIGMOD Record, Vol. 32, No. 2, pp. 5-14, [9] D. Abad, D. Carney, U. Cetntemel, M. Chernack, C. Convey, S. Lee, M. Stonebraker, N. Tatbul, and S. Zdonk, Aurora: A New Model and Archtecture for Data Stream Management, VLDB Journal, Vol. 12, No. 2, pp , [10] S. Chandraekaran, O. Cooper, A. Dehpande, M. J. Frankln, J. M. Hellerten, W. Hong, S. Krhnamurthy, S. R. Madden, F. Re, and M. A. Shah, TelegraphCQ: Contnuou Dataflow Proceng, Proc. ACM SIGMOD Int. Conf. on Management of Data, [11] I. Foter, Y. Zhao, I. Racu, S. Lu, Cloud Computng and Grd Computng 360-Degree Compared, Proc. IEEE Grd Computng Envronment, conj. IEEE/ACM Supercomputng Conf., Autn, [12] R. Buyya, C. S. Yeo, S. Venugopala, J. Broberga, and I. Brandcc, Cloud Computng and Emergng IT Platform: Von, Hype, and Realty for Delverng Computng a the 5th Utlty, Future Generaton Computer Sytem, Vol. 25, No. 6, pp , [13] M. Roenblum and T. Garfnkel, Vrtual Machne Montor: Current Technology and Future Trend, IEEE Computer, Vol. 38, No. 5, pp , [14] I. Foter, T. Freeman, K. Keahey, D. Scheftner, B. Sotomayer and X. Zhang, Vrtual Cluter for Grd Communte, Proc. IEEE Int. Symp. on Cluter Computng and the Grd, May [15] F. Zhang, J. Cao, H. Ca, L. Lu, and C. Wu, Redundant Vrtual Machne Management n Vrtualzed Cloud Platform, Int. J. Modelng, Smulaton, and Scentfc Computng, Vol. 2, No. 2, pp , [16] D. Nurm, R. Wolk, C. Grzegorczyk, G. Obertell, S. Soman, L. Youeff, and D. Zagorodnov, The Eucalyptu Open-Source Cloud-Computng Sytem, Proc. 9th IEEE/ACM Int. Symp. on Cluter Computng and the Grd, [17] I. Krul, A. Ganguly, J. Zhang, J. A. B. Forte, and R. J. Fgueredo, VMPlant: Provdng and Managng Vrtual Machne Executon Envronment for Grd Computng, Proc. ACM/IEEE SC2004 Conference, [18] M. McNett, D. Gupta, A. Vahdat, and G. M. Voelker, Uher: an Extenble Framework for Managng Cluter of Vrtual Machne, Proc. 21t Conf. on Large Intallaton Sytem Admntraton, [19] L. Chen and G. Agrawal, A Statc Reource Allocaton Framework for Grd-baed Streamng Applcaton, Concurrency and Computaton: Practce and Experence, 18: , [20] B. Agarwalla, N. Ahmed, D. Hlley, and U. Ramachandran, Streamlne: Schedulng Streamng Applcaton n a Wde Area Envronment, Multmeda Sytem, Vol. 13, No. 1, pp , [21] L. Battetll, et al., EnLIGHTened Computng: An Archtecture for Co-allocatng Network, Compute, and Other Grd Reource for Hgh-End Applcaton, Proc. Int. Symp. on Hgh Capacty Optcal Network and Enablng Technologe, pp. 1-8, [22] A. Takefua, et al., G-lambda: Coordnaton of a Grd Scheduler and Lambda Path Servce over GMPLS, Future Generaton Computer Sytem, Vol. 22, No. 8, pp , October [23] S. Fguerola, et al., PHOSPHORUS: Sngle-tep On-demand Servce acro Mult-doman Network for e-scence, Proc. SPIE, Vol. 6784, 67842X, [24] T. F. Abdelzaher, K. G. Shn, and N. Bhatt, Performance Guarantee for Web Server End-ytem: A Control-theoretcal Approach, IEEE Tran. on Parallel and Dtrbuted Sytem, Vol. 13, [25] S. Parekh, N. Gandh, J. L. Hellerten, D. Tlbury, T. S. Jayram, and J. Bgu, Ung Control Theory to Acheve Servce Level Objectve n Performance Management, Real Tme Sytem Journal, Vol. 23, No. 1-2, [26] Y. Dao, N. Gandh, J. L. Hellerten, S. Parekh, and D.M. Tlbury, MIMO Control of an Apache Web Server: Modelng and Controller Degn, Proc. Amercan Control Conf., [27] A. Kamra, V. Mra, and E. M. Nahum, Yakha: A Self-tunng Controller for Managng the Performance of 3-tered Web Ste, Proc. 12th IEEE Int. Workhop on Qualty of Servce, June, 2004.

12 > REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 12 [28] Y. Lu, C. Lu, T. Abdelzaher, and G. Tao, An Adaptve Control Framework for QoS Guarantee and t Applcaton to Dfferentated Cachng Servce, Proc. 10th IEEE Int. Workhop on Qualty of Servce, May [29] P. J. Kng and E. H. Mamdan, Applcaton of Fuzzy Algorthm for Control Smple Dynamc Plant, IEE Proc. Control Theory App., Vol. 121, pp , [30] Y. Dao, J. L. Hellerten, S. Parekh, Ung Fuzzy Control to Maxmze Proft n Servce Level Management, IBM Sytem Journal, Vol. 41, No. 3, [31] B. L and K. Nahrtedt, A Control-baed Mddleware Framework for Qualty of Servce Adaptaton, Communcaton of ACM, Vol. 17, pp , [32] M. H. Suzer and K. D. Kang, Adaptve Fuzzy Control for Utlzaton Management, Proc. IEEE Int. Symp. on Object/Component/ Servce-orented Real-tme Dtrbuted Computng, [33] S. Park and M. Humphrey, Feedback-controlled Reource Sharng for Predctable escence, Proc. ACM/IEEE Conf. on Supercomputng [34] R. Prodan, M. Weczorek, B-Crtera Schedulng of Scentfc Grd Workflow, IEEE Tranacton on Automaton Scence and Engneerng, Vol. 7, No. 2, pp , [35] W. Tan, Y. Fan, M. Zhou, and Z. Tan, Data-drven Servce Compoton n Buldng SOA Soluton: A Petr Net Approach, IEEE Tranacton on Automaton Scence and Engneerng, Vol. 7, No. 3, pp , [36] The Xen Hypervor. [37] W. Pedrycz Fuzzy Control and Fuzzy Sytem (2nd ed.). John Wley & Son, Inc., New York, NY, USA. [38] J. Cao and J. L, Real-tme Gravtatonal-wave Burt Search for Mult-meenger Atronomy, Int. J. Modern Phyc D, Vol. 20, No. 10, pp , [39] S. Pandey, Schedulng and Management of Data Intenve Applcaton Workflow n Grd and Cloud Computng Envronment, Doctor of Phloophy The, The Unverty of Melbourne, Autrala, [40] C. Herath, and B. Plale, Streamflow Programmng Model for Data Streamng n Scentfc Workflow, Proc. 10 th IEEE/ACM Int. Conf. on Cluter, Cloud and Grd Computng, pp , [41] C. S. Lew, M. P. Atknon, J. I. v. Hemert et al., Toward Optmng Dtrbuted Data Streamng Graph ung Parallel Stream, Proc. 19th ACM Int. Symp. on Hgh Performance Dtrbuted Computng, Chcago, Illno, pp , [42] D. Znn, Q. Hart, T. McPhllp et al., Toward Relable, Performant Workflow for Streamng-Applcaton on Cloud Platform, Proc. 11 th IEEE/ACM Int. Symp. on Cluter, Cloud and Grd Computng, pp , [43] J. Cao, E. Katavound, and J. Zwezg, Grd Enabled LIGO Data Montorng, Proc. IEEE/ACM Supercomputng Conf., Seattle, WA, USA, [44] R. Porde for the Open Scence Grd Conortum, The Open Scence Grd, Proc. Computng n Hgh Energy and Nuclear Phyc Conf., Interlaken, Swtzerland, [45] W. Chen, J. Cao, and Z. L, Cutomzed Vrtual Machne for Software Provonng n Scentfc Cloud, Proc. 2nd Int. Conf. on Networkng and Dtrbuted Computng, Bejng, Chna, [46] LIGO Scentfc Collaboraton, A Gravtatonal Wave Obervatory Operatng beyond the Quantum Shot-noe Lmt, Nature Phyc, Advance Onlne Publcaton, [47] P. Xong, Z. Wang, S. Malkowk, Q. Wang, D. Jayanghe and C. Pu, "Economcal and Robut Provonng of N-Ter Cloud Workload: A Mult-level Control Approach," Proc. IEEE Internatonal Conference On Dtrbuted Computng Sytem (ICDCS), Mnneapol, Mnneota, USA, BIOGRAPHIES Junwe Cao (M 99 SM 05) receved h Ph.D. n computer cence from the Unverty of Warwck, Coventry, UK, n He receved h bachelor and mater degree n control theore and engneerng n 1998 and 1996, repectvely, both from Tnghua Unverty, Bejng, Chna. He currently a Profeor and Vce Drector, Reearch Inttute of Informaton Technology, Tnghua Unverty, Bejng, Chna. He alo wth Tnghua Natonal Laboratory for Informaton Scence and Technology, Bejng, Chna. He now a Vtng Scentt of MIT LIGO Laboratory. Before jonng Tnghua Unverty n 2006, he wa a Reearch Scentt at MIT LIGO Laboratory and NEC Laboratore Europe for about 5 year. He ha publhed over 130 paper and cted by nternatonal cholar for over 2200 tme. He the book edtor of Cybernfratructure Technologe and Applcaton, publhed by Nova Scence n H reearch focued on advanced computng technologe and applcaton. Prof. Cao a Senor Member of the IEEE Computer Socety and a Member of the ACM and CCF. He alo a member of LIGO Scentfc Collaboraton. Wen Zhang (M 07) work n Chongqng Mltary Delegate Bureau, General Armament Department of PLA. He receved h PhD n control engneerng and applcaton from Department of Automaton, Tnghua Unverty, Bejng, Chna, n H reearch wa focued on dynamc control of data treamng and proceng. We Tan receved the B.S. degree and the Ph.D. degree n Control Scence and Engneerng, from Department of Automaton, Tnghua Unverty, Chna n 2002 and 2008, repectvely. He currently a reearch taff member n IBM T. J. Waton Reearch Center, USA. From 2008 to 2010 he wa a reearcher at Computaton Inttute, Unverty of Chcago and Argonne Natonal Laboratory, USA. H reearch nteret nclude ervce-orented archtecture, bune and centfc workflow, Petr net, workload optmzaton, bune ntellgence and data centrc computng. He ha publhed over 30 journal and conference paper. He Aocate Edtor of the IEEE Tranacton on Automaton Scence and Engneerng. He erved n program commttee of many nternatonal conference and wa the PC co-char of the Frt IEEE/ACM Workhop on the applcaton of Socal Networkng concept to Cluter, Cloud, Grd and Servce Computng (SN4CCGrdS). Contact hm at wtan@u.bm.com.

An Integrated Resource Management and Scheduling System for Grid Data Streaming Applications

An Integrated Resource Management and Scheduling System for Grid Data Streaming Applications An Integrated eource Management and Schedulng Sytem for Grd Data Streamng Applcaton Wen Zhang, Junwe Cao 2,3*, Yheng Zhong,3, Lanchen Lu,3, and Cheng Wu,3 Department of Automaton, Tnghua Unverty, Bejng

More information

PERFORMANCE ANALYSIS OF PARALLEL ALGORITHMS

PERFORMANCE ANALYSIS OF PARALLEL ALGORITHMS Software Analye PERFORMANCE ANALYSIS OF PARALLEL ALGORIHMS Felcan ALECU PhD, Unverty Lecturer, Economc Informatc Deartment, Academy of Economc Stude, Bucharet, Romana E-mal: alecu.felcan@e.ae.ro Abtract:

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

Multifunction Phased Array Radar Resource Management: Real-Time Scheduling Algorithm

Multifunction Phased Array Radar Resource Management: Real-Time Scheduling Algorithm Journal of Computatonal Informaton Sytem 7:2 (211) 385-393 Avalable at http://www.jofc.com Multfuncton Phaed Array Radar Reource Management: Real-me Schedulng Algorm Janbn LU 1,, Hu XIAO 2, Zemn XI 1,

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

T1 Estimates SAT - 2006

T1 Estimates SAT - 2006 T1 Etmate SAT - 006 Tax and Lmoune Servce (TL) NAICS : 4853** by Javer Oyarzun BSMD Stattc Canada 007-1-1 1. Introducton 1.1 Ue of admntratve data Over the lat few year, Stattc Canada (STC) ha been accentuatng

More information

THE ANALYSIS AND OPTIMIZATION OF SURVIVABILITY OF MPLS NETWORKS. Mohammadreza Mossavari, Yurii Zaychenko

THE ANALYSIS AND OPTIMIZATION OF SURVIVABILITY OF MPLS NETWORKS. Mohammadreza Mossavari, Yurii Zaychenko Internatonal Journal "Informaton Theore & Applcaton" Vol5 / 28 253 TE ANALYSIS AND OTIMIATION OF SURVIVABILITY OF MLS NETWORS Mohammadreza Moavar, Yur aychenko Abtract: The problem of MLS network urvvablty

More information

A Novel Architecture Design of Large-Scale Distributed Object Storage System

A Novel Architecture Design of Large-Scale Distributed Object Storage System Internatonal Journal of Grd Dtrbuton Computng Vol.8, No.1 (2015), pp.25-32 http://dx.do.org/10.14257/gdc.2015.8.1.03 A Novel Archtecture Degn of Large-Scale Dtrbuted Obect Storage Sytem Shan Yng 1 and

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

Study on Model of Risks Assessment of Standard Operation in Rural Power Network

Study on Model of Risks Assessment of Standard Operation in Rural Power Network Study on Model of Rsks Assessment of Standard Operaton n Rural Power Network Qngj L 1, Tao Yang 2 1 Qngj L, College of Informaton and Electrcal Engneerng, Shenyang Agrculture Unversty, Shenyang 110866,

More information

ITS-90 FORMULATIONS FOR VAPOR PRESSURE, FROSTPOINT TEMPERATURE, DEWPOINT TEMPERATURE, AND ENHANCEMENT FACTORS IN THE RANGE 100 TO +100 C.

ITS-90 FORMULATIONS FOR VAPOR PRESSURE, FROSTPOINT TEMPERATURE, DEWPOINT TEMPERATURE, AND ENHANCEMENT FACTORS IN THE RANGE 100 TO +100 C. ITS-90 FORMULATIONS FOR VAPOR PRESSURE, FROSTPOINT TEMPERATURE, DEWPOINT TEMPERATURE, AND ENHANCEMENT FACTORS IN THE RANGE 100 TO +100 C Bob Hardy Thunder Scentfc Corporaton, Albuquerque, NM, USA Abtract:

More information

Development and use of prediction models in Building Acoustics as in EN 12354. 1 Introduction. 2 EN 12354, part 1 & 2. 2.2 Lightweight single elements

Development and use of prediction models in Building Acoustics as in EN 12354. 1 Introduction. 2 EN 12354, part 1 & 2. 2.2 Lightweight single elements evelopment and ue of predcton model n Buldng Acoutc a n EN 1354 Eddy TNO Scence and Indutry, P.O. Box 155, N-600 A elft, The Netherland, eddy.gerreten@tno.nl Improvng the acoutc clmate n buldng an mportant

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

Control and Coordination of Interactive Videoconferencing over Hybrid Networks

Control and Coordination of Interactive Videoconferencing over Hybrid Networks 1 of 5 ontrol and oordnaton of Interactve Vdeoconferencng over Hybrd Network Tng-hao Hou, horng-horng Yang., Yun-Sun hu, and Km-Joan hen epartment of Electrcal Engneerng Natonal hung heng Unverty 160,

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

ARTICLE IN PRESS. JID:COMAID AID:1153 /FLA [m3g; v 1.79; Prn:21/02/2009; 14:10] P.1 (1-13) Computer Aided Geometric Design ( )

ARTICLE IN PRESS. JID:COMAID AID:1153 /FLA [m3g; v 1.79; Prn:21/02/2009; 14:10] P.1 (1-13) Computer Aided Geometric Design ( ) COMAID:5 JID:COMAID AID:5 /FLA [mg; v 79; Prn:/0/009; 4:0] P -) Computer Aded Geometrc Degn ) Content lt avalable at ScenceDrect Computer Aded Geometrc Degn wwwelevercom/locate/cagd Fat approach for computng

More information

A New Task Scheduling Algorithm Based on Improved Genetic Algorithm

A New Task Scheduling Algorithm Based on Improved Genetic Algorithm A New Task Schedulng Algorthm Based on Improved Genetc Algorthm n Cloud Computng Envronment Congcong Xong, Long Feng, Lxan Chen A New Task Schedulng Algorthm Based on Improved Genetc Algorthm n Cloud Computng

More information

The issue of whether the Internet will permanently destroy the news media is currently a

The issue of whether the Internet will permanently destroy the news media is currently a Wll the Internet etroy the New Meda? or Can Onlne Advertng Market Save the Meda? by Suan Athey, Emlo Calvano and Johua S. Gan * Frt raft: October, 009 Th Veron: November, 00 PRELIMINARY PLEASE O NOT QUOTE

More information

IMPACT ANALYSIS OF A CELLULAR PHONE

IMPACT ANALYSIS OF A CELLULAR PHONE 4 th ASA & μeta Internatonal Conference IMPACT AALYSIS OF A CELLULAR PHOE We Lu, 2 Hongy L Bejng FEAonlne Engneerng Co.,Ltd. Bejng, Chna ABSTRACT Drop test smulaton plays an mportant role n nvestgatng

More information

Rate Monotonic (RM) Disadvantages of cyclic. TDDB47 Real Time Systems. Lecture 2: RM & EDF. Priority-based scheduling. States of a process

Rate Monotonic (RM) Disadvantages of cyclic. TDDB47 Real Time Systems. Lecture 2: RM & EDF. Priority-based scheduling. States of a process Dsadvantages of cyclc TDDB47 Real Tme Systems Manual scheduler constructon Cannot deal wth any runtme changes What happens f we add a task to the set? Real-Tme Systems Laboratory Department of Computer

More information

DBA-VM: Dynamic Bandwidth Allocator for Virtual Machines

DBA-VM: Dynamic Bandwidth Allocator for Virtual Machines DBA-VM: Dynamc Bandwdth Allocator for Vrtual Machnes Ahmed Amamou, Manel Bourguba, Kamel Haddadou and Guy Pujolle LIP6, Perre & Mare Cure Unversty, 4 Place Jusseu 755 Pars, France Gand SAS, 65 Boulevard

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

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

ESSAYS IN RENEWABLE ENERGY AND EMISSIONS TRADING

ESSAYS IN RENEWABLE ENERGY AND EMISSIONS TRADING ESSAYS IN RENEWABLE ENERGY AND EMISSIONS TRADING By JOSHUA D. KNEIFEL A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE

More information

Modeling ISP Tier Design

Modeling ISP Tier Design Modelng ISP Ter Degn We Da School of Informaton and Computer Scence Unverty of Calforna, Irvne Irvne, CA, US daw1@uc.edu Scott Jordan School of Informaton and Computer Scence Unverty of Calforna, Irvne

More information

Mall Cell Network - Power, Memory and Networking

Mall Cell Network - Power, Memory and Networking Slow admon and power control for mall cell network va dtrbuted optmzaton Sew Eng Na, Tony Q.S. Quek, Mérouane Debbah To cte th veron: Sew Eng Na, Tony Q.S. Quek, Mérouane Debbah. Slow admon and power control

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

An Optimal Model for Priority based Service Scheduling Policy for Cloud Computing Environment

An Optimal Model for Priority based Service Scheduling Policy for Cloud Computing Environment An Optmal Model for Prorty based Servce Schedulng Polcy for Cloud Computng Envronment Dr. M. Dakshayn Dept. of ISE, BMS College of Engneerng, Bangalore, Inda. Dr. H. S. Guruprasad Dept. of ISE, BMS College

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

A heuristic task deployment approach for load balancing

A heuristic task deployment approach for load balancing Xu Gaochao, Dong Yunmeng, Fu Xaodog, Dng Yan, Lu Peng, Zhao Ja Abstract A heurstc task deployment approach for load balancng Gaochao Xu, Yunmeng Dong, Xaodong Fu, Yan Dng, Peng Lu, Ja Zhao * College of

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

REVISTA INVESTIGACIÓN OPERACIONAL VOL., 33, NO. 3, 233-244, 2012.

REVISTA INVESTIGACIÓN OPERACIONAL VOL., 33, NO. 3, 233-244, 2012. REVISA INVESIGACIÓN OPERACIONAL VOL., 33, NO. 3, 33-44,. ORDERING POLICY FOR INVENORY MANAGEMEN WHEN DEMAND IS SOCK- DEPENDEN AND A EMPORARY PRICE DISCOUN IS LINKED O ORDER UANIY Nta H. Shah Department

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

Basic Principle of Buck-Boost

Basic Principle of Buck-Boost Bac Prncple of Buck-Boot he buck-boot a popular non-olated nvertng power tage topology, ometme called a tep-up/down power tage. Power upply degner chooe the buck-boot power tage becaue the requred output

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

IWFMS: An Internal Workflow Management System/Optimizer for Hadoop

IWFMS: An Internal Workflow Management System/Optimizer for Hadoop IWFMS: An Internal Workflow Management System/Optmzer for Hadoop Lan Lu, Yao Shen Department of Computer Scence and Engneerng Shangha JaoTong Unversty Shangha, Chna lustrve@gmal.com, yshen@cs.sjtu.edu.cn

More information

How To Understand Propect Theory And Mean Variance Analysis

How To Understand Propect Theory And Mean Variance Analysis Invetment Management and Fnancal Innovaton, Volume 6, Iue 1, 2009 Enrco De Gorg (Swtzerland ), Thorten Hen (Swtzerland) Propect theory and mean-varance analy: doe t make a dfference n wealth management?

More information

Pass by Reference vs. Pass by Value

Pass by Reference vs. Pass by Value Pa by Reference v. Pa by Value Mot method are paed argument when they are called. An argument may be a contant or a varable. For example, n the expreon Math.qrt(33) the contant 33 paed to the qrt() method

More information

Abstract. 2.2. Adjusted PPM.

Abstract. 2.2. Adjusted PPM. Effectvene of Avance an Authentcate Packet Markng Scheme for Traceback of Denal of Servce Attack Blal Rzv an Emmanuel Fernánez-Gaucheran Department of Electrcal & Computer Engneerng &Computer Scence Unverty

More information

RESEARCH ON DUAL-SHAKER SINE VIBRATION CONTROL. Yaoqi FENG 1, Hanping QIU 1. China Academy of Space Technology (CAST) yaoqi.feng@yahoo.

RESEARCH ON DUAL-SHAKER SINE VIBRATION CONTROL. Yaoqi FENG 1, Hanping QIU 1. China Academy of Space Technology (CAST) yaoqi.feng@yahoo. ICSV4 Carns Australa 9- July, 007 RESEARCH ON DUAL-SHAKER SINE VIBRATION CONTROL Yaoq FENG, Hanpng QIU Dynamc Test Laboratory, BISEE Chna Academy of Space Technology (CAST) yaoq.feng@yahoo.com Abstract

More information

Lecture 2: Single Layer Perceptrons Kevin Swingler

Lecture 2: Single Layer Perceptrons Kevin Swingler Lecture 2: Sngle Layer Perceptrons Kevn Sngler kms@cs.str.ac.uk Recap: McCulloch-Ptts Neuron Ths vastly smplfed model of real neurons s also knon as a Threshold Logc Unt: W 2 A Y 3 n W n. A set of synapses

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

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

Efficient Bandwidth Management in Broadband Wireless Access Systems Using CAC-based Dynamic Pricing

Efficient Bandwidth Management in Broadband Wireless Access Systems Using CAC-based Dynamic Pricing Effcent Bandwdth Management n Broadband Wreless Access Systems Usng CAC-based Dynamc Prcng Bader Al-Manthar, Ndal Nasser 2, Najah Abu Al 3, Hossam Hassanen Telecommuncatons Research Laboratory School of

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

A Load-Balancing Algorithm for Cluster-based Multi-core Web Servers

A Load-Balancing Algorithm for Cluster-based Multi-core Web Servers Journal of Computatonal Informaton Systems 7: 13 (2011) 4740-4747 Avalable at http://www.jofcs.com A Load-Balancng Algorthm for Cluster-based Mult-core Web Servers Guohua YOU, Yng ZHAO College of Informaton

More information

Optimum Design of Magnetic Inductive Energy Harvester and its AC-DC Converter

Optimum Design of Magnetic Inductive Energy Harvester and its AC-DC Converter Otmum Degn of Magnetc nductve Energy Harveter and t AC-DC Converter Qan Sun, Sumeet Patl, Stehen Stoute, Nan-Xang Sun, Brad Lehman Deartment of Electrcal and Comuter Engneerng Northeatern Unverty Boton,

More information

LITERATURE REVIEW: VARIOUS PRIORITY BASED TASK SCHEDULING ALGORITHMS IN CLOUD COMPUTING

LITERATURE REVIEW: VARIOUS PRIORITY BASED TASK SCHEDULING ALGORITHMS IN CLOUD COMPUTING LITERATURE REVIEW: VARIOUS PRIORITY BASED TASK SCHEDULING ALGORITHMS IN CLOUD COMPUTING 1 MS. POOJA.P.VASANI, 2 MR. NISHANT.S. SANGHANI 1 M.Tech. [Software Systems] Student, Patel College of Scence and

More information

HP Mission-Critical Services

HP Mission-Critical Services HP Msson-Crtcal Servces Delverng busness value to IT Jelena Bratc Zarko Subotc TS Support tm Mart 2012, Podgorca 2010 Hewlett-Packard Development Company, L.P. The nformaton contaned heren s subject to

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

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

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

Overview of monitoring and evaluation

Overview of monitoring and evaluation 540 Toolkt to Combat Traffckng n Persons Tool 10.1 Overvew of montorng and evaluaton Overvew Ths tool brefly descrbes both montorng and evaluaton, and the dstncton between the two. What s montorng? Montorng

More information

Modeling and Analysis of 2D Service Differentiation on e-commerce Servers

Modeling and Analysis of 2D Service Differentiation on e-commerce Servers Modelng and Analyss of D Servce Dfferentaton on e-commerce Servers Xaobo Zhou, Unversty of Colorado, Colorado Sprng, CO zbo@cs.uccs.edu Janbn We and Cheng-Zhong Xu Wayne State Unversty, Detrot, Mchgan

More information

A Programming Model for the Cloud Platform

A Programming Model for the Cloud Platform Internatonal Journal of Advanced Scence and Technology A Programmng Model for the Cloud Platform Xaodong Lu School of Computer Engneerng and Scence Shangha Unversty, Shangha 200072, Chna luxaodongxht@qq.com

More information

Performance Analysis of Energy Consumption of Smartphone Running Mobile Hotspot Application

Performance Analysis of Energy Consumption of Smartphone Running Mobile Hotspot Application Internatonal Journal of mart Grd and lean Energy Performance Analyss of Energy onsumpton of martphone Runnng Moble Hotspot Applcaton Yun on hung a chool of Electronc Engneerng, oongsl Unversty, 511 angdo-dong,

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

The Impact of the Internet on Advertising Markets for News Media

The Impact of the Internet on Advertising Markets for News Media The Impact of the Internet on Advertng Market for New Meda by Suan Athey, Emlo Calvano and Johua S. Gan * Frt Draft: October, 009 Th Veron: October 0 In th paper, we explore the hypothe that an mportant

More information

Polarimetric parameters associated to commercial optical fibers

Polarimetric parameters associated to commercial optical fibers RESEARCH Revta Mexcana de Fíca 6 14 443 45 NOVEMBER-DECEMBER 14 Polarmetrc parameter aocated to commercal optcal fber O. J. Velae-Ecobar a, K. M. Sala-Alcántara b, R. Epnoa-Luna b,, G. Atondo-Rubo a, and

More information

New method for grain size characterization of a multi-crystalline silicon ingot

New method for grain size characterization of a multi-crystalline silicon ingot New method for gran ze characterzaton of a mult-crytallne lcon ngot Maxme Forter, Erwann Fourmond, Jean-Mare Lebrun, Roland Enhau, Jed Kraem, Mutapha Lemt To cte th veron: Maxme Forter, Erwann Fourmond,

More information

APPLICATION OF PROBE DATA COLLECTED VIA INFRARED BEACONS TO TRAFFIC MANEGEMENT

APPLICATION OF PROBE DATA COLLECTED VIA INFRARED BEACONS TO TRAFFIC MANEGEMENT APPLICATION OF PROBE DATA COLLECTED VIA INFRARED BEACONS TO TRAFFIC MANEGEMENT Toshhko Oda (1), Kochro Iwaoka (2) (1), (2) Infrastructure Systems Busness Unt, Panasonc System Networks Co., Ltd. Saedo-cho

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

Measuring adverse selection in managed health care

Measuring adverse selection in managed health care Ž. Journal of Health Economc 19 2000 829 854 www.elever.nlrlocatereconbae Meaurng advere electon n managed health care Rchard G. Frank a,), Jacob Glazer b, Thoma G. McGure c a HarÕard UnÕerty, HarÕard

More information

Coalition Formation for Sourcing Contract Design with Cooperative Replenishment in Supply Networks

Coalition Formation for Sourcing Contract Design with Cooperative Replenishment in Supply Networks Coalton Formaton for Sourcng Contract Degn wth Cooperatve Replenhment n Supply Network Shem Ben Jouda Saouen Krchen Wald Klb November 014 CIRRELT-014-61 Coalton Formaton for Sourcng Contract Degn wth Cooperatve

More information

Applying the Value/Petri Process to ERP Software Development in China

Applying the Value/Petri Process to ERP Software Development in China Applyng the Value/Petr Proce to ERP Software Development n Chna LGuo Huang Barry Boehm Computer Scence Department Unverty of Southern Calforna Lo Angele, CA 90089-0781, USA 001 213-740-6470 {lguohua, boehm}@

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

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

Cloud Auto-Scaling with Deadline and Budget Constraints

Cloud Auto-Scaling with Deadline and Budget Constraints Prelmnary verson. Fnal verson appears In Proceedngs of 11th ACM/IEEE Internatonal Conference on Grd Computng (Grd 21). Oct 25-28, 21. Brussels, Belgum. Cloud Auto-Scalng wth Deadlne and Budget Constrants

More information

Dynamic Pricing for Smart Grid with Reinforcement Learning

Dynamic Pricing for Smart Grid with Reinforcement Learning Dynamc Prcng for Smart Grd wth Renforcement Learnng Byung-Gook Km, Yu Zhang, Mhaela van der Schaar, and Jang-Won Lee Samsung Electroncs, Suwon, Korea Department of Electrcal Engneerng, UCLA, Los Angeles,

More information

Performance Evaluation of Infrastructure as Service Clouds with SLA Constraints

Performance Evaluation of Infrastructure as Service Clouds with SLA Constraints Performance Evaluaton of Infrastructure as Servce Clouds wth SLA Constrants Anuar Lezama Barquet, Andre Tchernykh, and Ramn Yahyapour 2 Computer Scence Department, CICESE Research Center, Ensenada, BC,

More information

The Design of Reliable Trust Management Systems for Electronic Trading Communities

The Design of Reliable Trust Management Systems for Electronic Trading Communities The Degn of Relale Trut Management Sytem for Electronc Tradng Communte Chryantho Dellaroca Sloan School of Management Maachuett Inttute of Technology Room E53-315 Camrdge, MA 02139 dell@mt.edu Atract:

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

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

Resource Scheduling Based on Dynamic Dependence Injection in Virtualization-based Simulation Grid

Resource Scheduling Based on Dynamic Dependence Injection in Virtualization-based Simulation Grid Proceedngs of the 200 4th Internatonal Conference on Computer Supported Cooperatve Work n Desgn Resource Schedulng Based on Dynamc Dependence Injecton n Vrtualzaton-based Smulaton Grd Hanbng Lu,Hongy Su,

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

Case Study: Load Balancing

Case Study: Load Balancing Case Study: Load Balancng Thursday, 01 June 2006 Bertol Marco A.A. 2005/2006 Dmensonamento degl mpant Informatc LoadBal - 1 Introducton Optmze the utlzaton of resources to reduce the user response tme

More information

Cost-based Scheduling of Scientific Workflow Applications on Utility Grids

Cost-based Scheduling of Scientific Workflow Applications on Utility Grids Cost-based Schedulng of Scentfc Workflow Applcatons on Utlty Grds Ja Yu, Rakumar Buyya and Chen Khong Tham Grd Computng and Dstrbuted Systems Laboratory Dept. of Computer Scence and Software Engneerng

More information

Advances in Military Technology Vol. 10, No. 1, June 2015

Advances in Military Technology Vol. 10, No. 1, June 2015 AM Avance n Mltary echnology Vol., No., June 5 Mechancal an Computatonal Degn for Control of a -PUS Parallel Robot-bae Laer Cuttng Machne R. Zavala-Yoé *, R. Ramírez-Menoza an J. Ruz-García ecnológco e

More information

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

M3S MULTIMEDIA MOBILITY MANAGEMENT AND LOAD BALANCING IN WIRELESS BROADCAST NETWORKS

M3S MULTIMEDIA MOBILITY MANAGEMENT AND LOAD BALANCING IN WIRELESS BROADCAST NETWORKS M3S MULTIMEDIA MOBILITY MANAGEMENT AND LOAD BALANCING IN WIRELESS BROADCAST NETWORKS Bogdan Cubotaru, Gabrel-Mro Muntean Performance Engneerng Laboratory, RINCE School of Electronc Engneerng Dubln Cty

More information

A Spam Message Filtering Method: focus on run time

A Spam Message Filtering Method: focus on run time , pp.29-33 http://dx.doi.org/10.14257/atl.2014.76.08 A Spam Meage Filtering Method: focu on run time Sin-Eon Kim 1, Jung-Tae Jo 2, Sang-Hyun Choi 3 1 Department of Information Security Management 2 Department

More information

Testing and Debugging Resource Allocation for Fault Detection and Removal Process

Testing and Debugging Resource Allocation for Fault Detection and Removal Process Internatonal Journal of New Computer Archtectures and ther Applcatons (IJNCAA) 4(4): 93-00 The Socety of Dgtal Informaton and Wreless Communcatons, 04 (ISSN: 0-9085) Testng and Debuggng Resource Allocaton

More information

Frequency Selective IQ Phase and IQ Amplitude Imbalance Adjustments for OFDM Direct Conversion Transmitters

Frequency Selective IQ Phase and IQ Amplitude Imbalance Adjustments for OFDM Direct Conversion Transmitters Frequency Selectve IQ Phase and IQ Ampltude Imbalance Adjustments for OFDM Drect Converson ransmtters Edmund Coersmeer, Ernst Zelnsk Noka, Meesmannstrasse 103, 44807 Bochum, Germany edmund.coersmeer@noka.com,

More information

Causal, Explanatory Forecasting. Analysis. Regression Analysis. Simple Linear Regression. Which is Independent? Forecasting

Causal, Explanatory Forecasting. Analysis. Regression Analysis. Simple Linear Regression. Which is Independent? Forecasting Causal, Explanatory Forecastng Assumes cause-and-effect relatonshp between system nputs and ts output Forecastng wth Regresson Analyss Rchard S. Barr Inputs System Cause + Effect Relatonshp The job of

More information

QoS-based Scheduling of Workflow Applications on Service Grids

QoS-based Scheduling of Workflow Applications on Service Grids QoS-based Schedulng of Workflow Applcatons on Servce Grds Ja Yu, Rakumar Buyya and Chen Khong Tham Grd Computng and Dstrbuted System Laboratory Dept. of Computer Scence and Software Engneerng The Unversty

More information

QoS-Aware Active Queue Management for Multimedia Services over the Internet

QoS-Aware Active Queue Management for Multimedia Services over the Internet QoS-Aware Actve Queue Management for Multmeda Servces over the Internet I-Shyan Hwang, *Bor-Junn Hwang, Pen-Mng Chang, Cheng-Yu Wang Abstract Recently, the multmeda servces such as IPTV, vdeo conference

More information

taposh_kuet20@yahoo.comcsedchan@cityu.edu.hk rajib_csedept@yahoo.co.uk, alam_shihabul@yahoo.com

taposh_kuet20@yahoo.comcsedchan@cityu.edu.hk rajib_csedept@yahoo.co.uk, alam_shihabul@yahoo.com G. G. Md. Nawaz Al 1,2, Rajb Chakraborty 2, Md. Shhabul Alam 2 and Edward Chan 1 1 Cty Unversty of Hong Kong, Hong Kong, Chna taposh_kuet20@yahoo.comcsedchan@ctyu.edu.hk 2 Khulna Unversty of Engneerng

More information

Research of Network System Reconfigurable Model Based on the Finite State Automation

Research of Network System Reconfigurable Model Based on the Finite State Automation JOURNAL OF NETWORKS, VOL., NO. 5, MAY 24 237 Research of Network System Reconfgurable Model Based on the Fnte State Automaton Shenghan Zhou and Wenbng Chang School of Relablty and System Engneerng, Behang

More information

Methodology to Determine Relationships between Performance Factors in Hadoop Cloud Computing Applications

Methodology to Determine Relationships between Performance Factors in Hadoop Cloud Computing Applications Methodology to Determne Relatonshps between Performance Factors n Hadoop Cloud Computng Applcatons Lus Eduardo Bautsta Vllalpando 1,2, Alan Aprl 1 and Alan Abran 1 1 Department of Software Engneerng and

More information

Optimization Model of Reliable Data Storage in Cloud Environment Using Genetic Algorithm

Optimization Model of Reliable Data Storage in Cloud Environment Using Genetic Algorithm Internatonal Journal of Grd Dstrbuton Computng, pp.175-190 http://dx.do.org/10.14257/gdc.2014.7.6.14 Optmzaton odel of Relable Data Storage n Cloud Envronment Usng Genetc Algorthm Feng Lu 1,2,3, Hatao

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

Fair Virtual Bandwidth Allocation Model in Virtual Data Centers

Fair Virtual Bandwidth Allocation Model in Virtual Data Centers Far Vrtual Bandwdth Allocaton Model n Vrtual Data Centers Yng Yuan, Cu-rong Wang, Cong Wang School of Informaton Scence and Engneerng ortheastern Unversty Shenyang, Chna School of Computer and Communcaton

More information

Robust biometric-based user authentication scheme for wireless sensor networks

Robust biometric-based user authentication scheme for wireless sensor networks Robut bometrc-baed uer autentcaton ceme for wrele enor network Debao He* cool of Matematc and tattc Wuan nverty Wuan Cna Emal: edebao@16.com Abtract: Wrele enor network (WN) are appled wdely a varety of

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

Performance Management and Evaluation Research to University Students

Performance Management and Evaluation Research to University Students 631 A publcaton of CHEMICAL ENGINEERING TRANSACTIONS VOL. 46, 2015 Guest Edtors: Peyu Ren, Yancang L, Hupng Song Copyrght 2015, AIDIC Servz S.r.l., ISBN 978-88-95608-37-2; ISSN 2283-9216 The Italan Assocaton

More information

Atkinson-Stiglitz and Ramsey reconciled: Pareto e cient taxation and pricing under a break-even constraint

Atkinson-Stiglitz and Ramsey reconciled: Pareto e cient taxation and pricing under a break-even constraint Abtract The Ramey tax problem examne the degn o lnear commodty taxe to collect a gven tax revenue Th approach ha been erouly challenged by Atknon and Stgltz (976) who how that (under ome condton) an optmal

More information

A Hierarchical Reliability Model of Service-Based Software System

A Hierarchical Reliability Model of Service-Based Software System 2009 33rd Annual IEEE Internatonal Computer Software and Applcatons Conference A Herarchcal Relablty Model of Servce-Based Software System Lun Wang, Xaoyng Ba, Lzhu Zhou Department of Computer Scence and

More information

How To Model A Multi-Home

How To Model A Multi-Home The Impact of the Internet on Advertng Market for New Meda by Suan Athey, Emlo Calvano and Johua S. Gan * Frt raft: October, 009 Th Veron: Aprl 03 In th paper, we explore the hypothe that an mportant force

More information

SPEE Recommended Evaluation Practice #6 Definition of Decline Curve Parameters Background:

SPEE Recommended Evaluation Practice #6 Definition of Decline Curve Parameters Background: SPEE Recommended Evaluaton Practce #6 efnton of eclne Curve Parameters Background: The producton hstores of ol and gas wells can be analyzed to estmate reserves and future ol and gas producton rates and

More information

Imperial College London

Imperial College London F. Fang 1, C.C. Pan 1, I.M. Navon 2, M.D. Pggott 1, G.J. Gorman 1, P.A. Allson 1 and A.J.H. Goddard 1 1 Appled Modellng and Computaton Group Department of Earth Scence and Engneerng Imperal College London,

More information