Performance Study of Parallel Programming on Cloud Computing Environments Using MapReduce

Size: px
Start display at page:

Download "Performance Study of Parallel Programming on Cloud Computing Environments Using MapReduce"

Transcription

1 Performance Study of Parallel Programmng on Cloud Computng Envronments Usng MapReduce Wen-Chung Shh, Shan-Shyong Tseng Department of Informaton Scence and Applcatons Asa Unversty Tachung, 41354, Tawan {wjshh, Chao-Tung Yang* Department of Computer Scence and Informaton Engneerng, Tungha Unversty Tachung, 40704, Tawan Abstract Dvsble load applcatons have such a rch source of parallelsm that ther parallelzaton can sgnfcantly reduce ther total completon tme on cloud computng envronments. However, t s a challenge for cloud users, probably scentsts and engneers, to develop ther applcatons whch can explot the computng power of the cloud. Usng MapReduce, novce cloud programmers can easly develop a hgh performance cloud applcaton. To examne the performance of programs developed by ths approach, we apply ths pattern to mplement three knds of applcatons and conduct experments on our cloud test-bed. Expermental results show that MapReduce programmng s sutable for regular workload applcatons. Keywords- hgh performance computng; parallel programmng; cloud computng; MapReduce; Hadoop I. INTRODUCTION In recent years, ntegraton of nexpensve commodty computers, such as computng clusters and grds, have become promsng alternatves to tradtonal multprocessors [1, 2]. Among these, cloud computng has emerged as the nextgeneraton hgh performance computng platform. Bascally, cloud platforms are dstrbuted systems whch share resources through the form of nternet servces. On the one hand, users can access more computng resources through cloud technologes wthout knowng low-level detals. On the other hand, cloud envronments requre effectve management to operate n an effcent way. Moreover, the heterogenety and dynamc changng of the cloud envronment make t dfferent from conventonal parallel and dstrbuted computng systems, such as multprocessors and computng clusters. Therefore, t s a challenge to utlze the cloud effcently. Applcatons wth dvsble loads are a rch source of parallelsm [30]. Programmers can dentfy ndependent work unts wthn a program and dspatch them to dfferent processors to reduce ts completon tme. Nowadays, parallelzng a program for cloud platforms manly depends on human efforts. Automatc transformaton of parallel applcatons nto grd-aware ones was nvestgated n [3-5], but ther approach s not sutable for a novce programmer to develop parallel applcatons from scratch. Furthermore, t s dffcult for programmers to acqure real-tme cloud status nformaton and to approprately dstrbute workload wthn a program to heterogeneous workng nodes. Our dea s to provde programmers wth a parallel programmng pattern, whch takes care of detals related to cloud nfrastructure. All the programmers need to do s to fll n the pattern algorthm wth applcaton-specfc code fragments. The resultng program can approprately dstrbute the workload of the program to workng nodes accordng to dynamc node performance. That s, we propose a performance-based pattern algorthm, whch serves as a template for programmers to develop a parallel program on cloud platforms. To verfy ths approach, we have mplemented ths pattern usng Hadoop MapReduce [31, 32] and appled ths pattern to three types of applcatons, Matrx Multplcaton, Assocaton Rule Mnng and Mandelbrot Set Computaton. Expermental results on a cloud test-bed show that programs developed by ths approach can explot the computng power of the cloud. The prmary advantage of ths approach s that a programmer can easly develop hgh performance programs to execute on cloud envronments. The hgh performance results from two features of ths pattern. Frst, t s a hybrd method. In ts frst phase, workload s dstrbuted statcally accordng to node performance to reduce schedulng overhead. In the second phase, the remanng load s dspatched dynamcally to acheve load balance. Second, t utlzes real-tme nformaton to estmate the performance of the cloud. The pattern acqures cloud status nformaton from a montorng tool and estmates the performance of computng and communcaton resources wth the nformaton. Our contrbutons can be summarzed as follows. Frst, ths paper proposes a performance-based pattern for programmers to develop hgh-qualty parallel applcatons wth ease. Programs developed by ths approach can utlze cloud nformaton to adaptvely dstrbute workloads wthn a program. Second, we apply ths pattern to three knds of dvsble load applcatons on our cloud test-bed. Consequently, expermental results show the obvous effectveness of our approach. Note that ths work ams at a general pattern of workload dstrbuton, nstead of proposng a new loop schedulng scheme or a novel data mnng algorthm. II. RELATED WORK In ths secton, the theory of dvsble load s brefly revewed. Then, we present some well-known loop schedulng *Correspondng author /10/$ IEEE

2 schemes. A. Dvsble Load Theory Dvsble Load Theory (DLT) addresses the case where the total workload can be parttoned nto any number of ndependent sub-jobs. In the past, the theory of dvsble load has been wdely nvestgated n statc heterogeneous systems. However, t has not been wdely appled to computng cloud platforms, whch are characterzed by heterogeneous resources and dynamc envronments. Ths problem has been dscussed n the past decade, and a good revew can be found n [6]. In [7, 8], an exact method for dvsble load was proposed, whch was not from a dynamc and pragmatc vewpont as ours. DLT focuses on coarse-gran loads, whch are a pool of jobs or programs. However, the target of ths work s fne-gran loads, whch mght be loop teratons wthn a program, for example. We focus on the problem of parallelzng an applcaton wth dvsble loads for rapd executon on cloud envronments. Snce cloud envronments are dynamcally changng and heterogeneous, the problem s obvously dfferent from the tradtonal DLT problem. B. Loop Schedulng Schemes Conventonally, loop schedulng schemes are classfed accordng to the tme when the schedulng decson s made. Statc loop schedulng schemes make a schedulng decson at comple tme, and equally assgn the total teratons of a loop to processors. It s appled when each teraton of a loop takes roughly the same amount of tme, and the compler knows enough related nformaton before complaton. Its advantage s less overhead at runtme, whle the dsadvantage s possble load mbalance. Well-known statc schedulng schemes nclude Block Schedulng, Cyclc Schedulng, Block-D Schedulng, Cyclc-D Schedulng, etc. However, these schemes are not sutable for dynamc grd envronments. Dynamc loop schedulng schemes make a schedulng decson at runtme. Its dsadvantage s more overhead at runtme, whle the advantage s load balance. Several selfschedulng schemes are restated here as follows. Pure Self-schedulng (PSS) Ths s a straghtforward dynamc loop schedulng algorthm [9]. Whenever a processor becomes dle, a loop teraton s assgned to t. Ths algorthm acheves good load balance but also nduces excessve overhead. Chunk Self-schedulng (CSS) Instead of assgnng one teraton to an dle processor at one tme, CSS assgns k teratons each tme, where k, called the chunk sze, s a constant. When the chunk sze s one, ths scheme s PSS, as dscussed above. If the chunk sze s set to the bound of the parallel loop equally dvded by the number of processors, ths scheme becomes statc schedulng. A large chunk sze wll cause load mbalance whle a small chunk sze s lkely to result n too much runtme overhead. Guded Self-schedulng (GSS) Ths scheme can dynamcally change the number of teratons assgned to each processor [10]. More specfcally, the next chunk sze s determned by dvdng the number of remanng teratons of a parallel loop by the number of avalable processors. The property of decreasng chunk sze mples an effort s made to acheve load balance and to reduce the runtme overhead. By assgnng large chunks at the begnnng of a parallel loop, one can reduce the frequency of communcaton between the master and slaves. Factorng Self-schedulng (FSS) In some cases, GSS mght assgn too much work to the frst few processors, so that the remanng teratons are not tme-consumng enough to balance the workload. The Factorng algorthm addresses ths problem [11]. The assgnment of loop teratons to workng processors proceeds n phases. Durng each phase, only a subset of the remanng loop teratons (usually half) s dvded equally among the avalable processors. Therefore, t balances loads better than GSS does when the computaton tmes of loop teratons vary substantally. In addton, the synchronzaton overhead of Factorng s not sgnfcantly larger than that of GSS. Trapezod Self-schedulng (TSS) Ths approach tres to reduce the need for synchronzaton whle stll mantanng a reasonable load balance [12]. TSS(N s, N f ) assgns the frst N s teratons of a loop to the processor startng the loop and the last N f teratons to the processor performng the last fetch, where N s and N f are both specfed by the programmer or the system. Ths algorthm allocates large chunks of teratons to the frst few processors and successvely smaller chunks to the last few processors. Tzen and N proposed TSS(N/2p, 1) as a general selecton. In [13], the authors enhanced well-known loop selfschedulng schemes to ft an extremely heterogeneous PC cluster envronment. A two-phased approach was proposed to partton loop teratons and t acheved good performance n heterogeneous test-beds. For example, GSS can be enhanced by parttonng percent of the total teratons accordng to ther performance weghted by CPU clock n the frst phase. Then, the remander of the workload s stll scheduled by GSS. Ths enhanced scheme s called NGSS. In [14], NGSS was further enhanced by dynamcally adjustng the parameter accordng to system heterogenety. A performance benchmark was used to determne whether target systems are relatvely homogeneous or relatvely heterogeneous. In addton, the types of loop teratons were classfed nto four classes, and were analyzed respectvely. The scheme enhanced from GSS s called ANGSS. Our prevous work [15, 16] presents dfferent heurstcs to the parallel loop self-schedulng problem. Ths paper extends the dea of performance-based schedulng to desgn a performance-based skeleton for developng hgh performance applcatons on grds. Ths approach s appled to both the parallel loop self-schedulng applcaton and the assocaton rule mnng applcaton. In [30], the dea of performancebased schedulng was extended to desgn a performance-based skeleton for developng hgh performance applcatons on grd platforms. Ths work dffers from the above-mentoned researches n that the computng envronment s focused on cloud platforms.

3 III. APPROACH In ths secton, the MapReduce model s ntroduced frst. Then, the concepts of performance rato and statc-workload rato are revewed. Fnally, we present the programmng pattern for the performance-based cloud computng. A. MapReduce Model The MapReduce programmng model can be used to process large-scale data sets n cloud envronments. It conssts of three types of nodes: Master, Mappers and Reducer, as shown n Fgure 1. The Master dspatches sub-jobs to a set of Mappers. After these Mappers complete ther assgned jobs, the results are merged by the Reducer. Fgure 1. Overvew of the MapReduce model Hadoop [31] mplements the MapReduce model, whch provdes a hgh-level vew to parallel programmers. Wth the help of Hadoop, programmers can focus on hgh-level operatons. In the MapReduce model, the Mappers get subjobs n the form of <key, value>. Takng a large data set as example, a sub-job can be represented by < key, value >, where key s the flename of the data subset, and value s the content of the data subset. B. Performance Rato The concept of performance rato was prevously defned n [15, 16] n dfferent forms and parameters, accordng to the requrements of applcatons. In ths work, the pattern algorthm uses a performance functon to model the heterogeneous performance of the dynamc cloud nodes. The purpose of calculatng performance rato s to estmate the current processng capablty for each node. Wth ths metrc, the program can dstrbute approprate workloads to each node, and load balance can be acheved. The more accurate the estmaton s, the better the load balance s. Assume that m s the number of attrbutes. For example, ths study adopts three attrbutes: CPU speed, CPU loadng, and Bandwdth. Therefore, m s equal to 3. To estmate the performance of each slave node, a performance functon (PF) s defned for a slave node j, as shown n (1): PF j (V 1, V 2,, V m ) (1) where V, 1< <m, s a varable of the performance functon. In more detal, the varables could nclude CPU speed, networkng bandwdth, memory sze, etc. We propose to utlze a Cloud resource montorng tool, Hadoop, to acqure the values of attrbutes for all slaves. The PF for node j s defned as (2). PF j = CS j CL j CS CL node N where N s the set of all avalable cloud nodes. CS s the CPU clock speed of node, and t s a constant attrbute. The value of ths parameter s acqured by the montorng tool. CL s the CPU loadng of node, and t s a varable attrbute. The value of ths parameter s acqured by the montorng tool. The performance rato (PR) s defned to be the rato of all performance functons. For nstance, assume the PF values of three nodes are 1/2, 1/3 and 1/4. Then, the PR s 1/2:1/3:1/4;.e., the PR of the three nodes s 6:4:3. In other words, f there are 13 loop teratons, 6 teratons wll be assgned to the frst node, 4 teratons wll be assgned to the second node, and 3 teratons wll be assgned to the last one. C. Determnaton of Statc-Workload Rato (SWR) Another mportant factor to be estmated s the varaton degree among all unts of workloads. The concept of Statc- Workload Rato (SWR) was prevously defned n [30]. For example, Mandelbrot Set Computaton s a problem nvolvng rregular workloads. In each teraton of a loop, the workload s dfferent and vares sgnfcantly, as shown n Fgure 2. Obvously, a dstrbuton scheme whch does not consder the effect of rregular workload could not estmate PR accurately. No. of Iteratons of Y th teraton of X Fgure 2. The Mandelbrot Set on [-1.8, 0.5] to [-1.2, 1.2] We propose to use a parameter, SWR (Statc-Workload Rato), rangng from 0 to 1, to estmate the proporton of the workload whch can be statcally scheduled, allevatng the effect of rregular workload. In order to take advantage of statc schedulng, the SWR proporton of the total workload s dspatched accordng to Performance Rato. The desgn ratonale s based on a conservatve heurstc to estmate the rregular degree of workloads among all teratons. If the workload of the target applcaton s regular, SWR can be set to be 1. However, f the applcaton has rregular workload, such as Mandelbrot Set Computaton, t s reasonable to (2)

4 reserve some amount of workload for load balancng. We propose to randomly take fve samplng teratons, and compute ther executon tme. Then, the SWR value for the target applcaton s determned by (3). mn SWR = (3) MAX where mn s the mnmum executon tme of all sampled teratons for applcaton. MAX s the maxmum executon tme of all sampled teratons for applcaton. For example, for a regular applcaton wth unform workload dstrbuton, the fve sampled teratons are the same. Therefore, the SWR s 1, and the whole workload can be dspatched accordng to Performance Rato, wth good load balance. However, for another applcaton, the fve samplng executon tme mght be 7, 7.5, 8, 8.5 and 10 seconds, respectvely. Then the SWR s 7/10. Therefore, 70% of the workload would be scheduled statcally accordng to PR, whle 30% of the workload would be scheduled by a dynamc scheme. Module Reduce Intalzaton Whle (a chunk of workload arrves) { receve the chunk of workload Compute on ths chunk Send the result to the Master } Fnalzaton END Reduce IV. EXPERIMENTAL RESULTS To verfy our approach, a cloud test-bed was bult, and three types of applcaton programs were mplemented usng the Hadoop pg language: Matrx Multplcaton, Assocaton Rule Mnng and Mandelbrot Set Computaton. The former two applcatons have regular workloads, whle the last has rregular workload. A. Cloud Test-bed A Cloud computng test-bed has been bult by the Hgh Performance Computng Lab. of Tungha Unversty, Tawan [33], usng Hadoop. The summary of the nodes s shown n Fgure 3. Fgure 4 shows the real-tme status of the cloud testbed acqured by the montorng tool. D. Programmng Pattern Based on the estmated nformaton of workload dstrbuton and node performance, we propose an MapReduce programmng pattern for performance-based workload dstrbuton on cloud envronments. Ths pattern conssts of two modules: a Map module and a Reduce module. The Map module makes the schedulng decson and dspatches workloads to slaves. On the other hand, the Reduce module processes the assgned work. Ths algorthm s just a pattern, and the detaled mplementaton, such as data preparaton, parameter passng, etc., mght be dfferent accordng to requrements of varous applcatons. Intalzaton Module Map /* Stage 1: Gatherng the nformaton */ collect the followng nformaton: CPU_Loadng CPU_Clock_Speed collect the executon tme of 5 sampled teratons /* Stage 2: Calculate schedulng parameters */ calculate SWR of the workload calculate Performance Rato of all slave nodes Fgure 3. The node summary of the cloud test-bed /* Stage 3: Statc Schedulng */ dspatch the (SWR)-percent of workload accordng to Performance Rato probe and receve for returned results /* Stage 4: dynamc Schedulng */ dspatch the (100-SWR)-percent of workload by a dynamc scheme Fnalzaton END Map

5 Fgure 4. The snapshot of the montorng tool on the TIGER Cloud In ths study, we have mplemented several schedulng schemes for the purpose of evaluaton. The conventonal statc schedulng scheme s to equally dstrbute the total workload to each worker at comple tme. However, ths scheme s obvously not sutable for dynamc and heterogeneous envronments. Therefore, a weghted statc schedulng scheme s adopted n ths experment. The prncple of parttonng s accordng to the CPU clock speed of each processor. A faster node wll get more workloads than a slower one proportonally. To reduce errors of expermental results, executon tme n each experment s obtaned by averagng the results of fve repettve executons. B. Applcaton 1: Matrx Multplcaton Matrx Multplcaton s a fundamental operaton n many numercal lnear algebra applcatons. Its effcent mplementaton on parallel computers s an ssue of prme mportance when provdng such systems wth scentfc software lbrares. Consequently, consderable effort has been devoted n the past to the development of effcent parallel matrx multplcaton algorthms, and ths wll reman a task n the future as well. Many parallel algorthms have been desgned, mplemented, and tested on dfferent parallel computers or cluster of workstatons for matrx multplcaton. In ths applcaton, the workload s loop teratons. The Master module s responsble for the dstrbuton of workloads. When a slave node becomes dle, the master node sends two ntegers to the slave. The two numbers represent the begnnng and endng ponters to the assgned chunk respectvely. In other words, every node has a copy of the nput matrces locally, so data communcaton s not sgnfcant n ths knd of mplementaton. Therefore, communcaton cost between the master and the slave s low, and the domnant cost s the computaton of matrx multplcaton. Frst, we want to compare the proposed PWD scheme wth prevous schemes wth respect to the executon tme. Fgure 5 llustrates the executon tme of weghted statc schedulng, GSS, FSS, TSS, NGSS, ANGSS and our PWD scheme, wth nput matrx sze , , and respectvely. The results are shown as follows. Among these schemes, PWD performs better than other schemes. The reason s that PWD accurately estmates the PR, and takes the advantage of statc schedulng, thus reducng the runtme overhead. The statc scheme obvously performs worse than other dynamc schemes. It s reasonable to say that the statc scheme s not sutable for a dynamc cloud envronment, wth respect to performance. It s nterestng that tradtonal self-schedulng schemes (FSS and TSS) perform slghtly better than NGSS and ANGSS. However, ths result s nconsstent wth that of prevous research [13, 14]. The reason mght be that the parameter s set too hgh, 75. If the parameter s set approprately, t s possble for NGSS and ANGSS to perform better, as prevous work has shown. Ths case also ndcates that NGSS and ANGSS suffer from the dffculty of determnng an approprate parameter value. Fgure 5. Executon tme for Matrx multplcaton wth dfferent nput szes C. Applcaton 2: Assocaton Rule Mnng Data mnng, or known as knowledge dscovery, s to acqure nterestng knowledge from large-scale databases [20]. Data mnng technques nclude assocaton rule mnng, classfcaton, cluster analyss, etc. The objectve of assocaton rule mnng s to dscover correlaton relatonshps among a set of tems. The well-known applcaton of assocaton rule mnng s market basket analyss. Ths technque can extract customer buyng behavors by dscoverng what tems they buy together. The managers of shops can place the assocated tems at the neghborng shelf to rase ther probablty of purchasng. For example, mlk and bread are frequently bought together. The formulaton of assocaton rule mnng problem s descrbed as follows [21, 22]. Let I be a set of tems, and D a database of transactons. Each transacton n D s a subset of I. An assocaton rule s a rule of the form A B, where A I, B I, and A B=. The well-known algorthm for fndng assocaton rules n large transacton databases s Apror. It utlzes the Apror property to reduce the search space. As the rsng of parallel processng, parallel data mnng have been well nvestgated n the past decade. Especally,

6 much attenton has been drected to parallel assocaton rule mnng. A good survey can be found n [23]. Tradtonal parallel data mnng work assumes data s parttoned and transmtted to the computng nodes n advance. However, t s usually the case n whch a large database s generated and stored n some staton. Therefore, t s mportant to effcently partton and to dstrbute the data to other nodes for parallel computaton. In ths applcaton, the workload s a database of transactons. We appled the pattern to mplement the Apror algorthm and ts data dstrbuton. Specfcally, the parallelzed verson of Apror we adopt s Count Dstrbuton (CD) [21, 22]. Our datasets are generated by the tool ndcated n [22]. The parameters of the synthetc datasets are descrbed n Table I. Dataset TABLE I. Number of Transactons DESCRIPTION OF OUR DATASET Average Transacton Length Number of Items D10KT5I10 10, D50KT5I10 50, D100KT5I10 100, D150KT5I10 150, Fgure 6 llustrates the executon tme of dfferent scheme, wth nput sze 10K, 50K, 100K and 150K transactons respectvely. Expermental results show that the scheme mplemented by the pattern got better performance than other schemes. From ths experment, we can see the sgnfcant nfluence of workload dstrbuton schemes on the total response tme. In cloud envronments, network bandwdth s an mportant crteron to evaluate the performance of a slave node. The Statc scheme can not adapt to the practcal network status. In contrast to Statc, when communcaton cost becomes a major factor, dynamc schemes would be well adaptve to the network envronment. Moreover, the reason why PWD got the best performance can be attrbuted to the approprate estmaton of node performance, especally for the attrbute of network bandwdth. In cloud computng envronments, CPU speed s not the only factor to determne the node performance. A node wth the fastest CPU s not necessary the node wth optmal performance. Fgure 6. Performance of data partton schemes for dfferent datasets D. Applcaton 3: Mandelbrot Set Computaton The Mandelbrot set computaton s a problem nvolvng the same computaton on dfferent data ponts whch have dfferent convergence rates [24]. Ths operaton derves a resultant mage by processng an nput matrx, A, where A s an mage of a pxels by b pxels. The resultant mage s one of a pxels by b pxels. The Mandelbrot Set Computaton has been mplemented usng the pattern. The Master module s responsble for the dstrbuton of workload. When a slave node becomes dle, the master node sends two ntegers to the slaves. As mplemented n Matrx Multplcaton, communcaton cost between the master and the slave s low, and the domnant cost s the computaton of the Mandelbrot Set. In the followng experment, we want to compare the executon tme of prevous schemes wth the mplemented program. Fgure 7 llustrates the executon tme of GSS, FSS, TSS, NGSS, ANGSS and our PWD scheme, wth nput mage sze 64 64, , and respectvely. The executon tme of weghted statc schedulng s omtted because ts results are sgnfcantly nferor to other schemes. Accordng to the experence n the Matrx Multplcaton applcaton, the parameter n NGSS s set to 30. The results are shown as follows. Among these schemes, PWD stll performs better than other schemes. The reason s also that PWD accurately estmates the PR, and takes the advantage of statc schedulng, thus reducng the runtme overhead. Tradtonal self-schedulng schemes (GSS, FSS and TSS) perform worse than NGSS and ANGSS. The reason s that t s dffcult to effcently schedule rregular workload for conventonal dynamc schemes. If the parameter s set approprately, t s certan for NGSS and ANGSS to perform better than GSS, FSS and TSS, as prevous work has shown.

7 applcatons for cloud envronments, whch s a dffcult task for novce programmers. Nevertheless, wth the pattern, all a programmer need to do s just to fll the applcaton-specfc program codes nto the pattern. If a programmer can code a sequental program, then t s straghtforward to transform t to a cloud applcaton. To extend the pattern dea to non-masterslave applcatons, such as P2P networks, we need to acqure experences and expertse n P2P programmng. In addton, the lack of global statstcal n non-master-slave applcatons s a problem to be solved. In P2P networks, the performancerelated nformaton can be gathered through socal actvtes, such as gossp protocols. Ths wll be an nterestng research topc n our future work. Fgure 7. Executon tme for Mandelbrot Set Computaton wth dfferent nput szes E. Dscusson In ths secton, several ssues are dscussed to clarfy the proposed approach. In general, task schedulng n cloud systems manly focuses on fne gran parallelsm, under the consderaton of the system heterogenety and the messagepassng communcaton. However, one goal of cloud computng s to explot potental parallelsm n nternet-scale cloud envronments. In addton to coarse gran parallelsm, we thnk that t s benefcal to explot fne gran parallelsm n cloud systems. The frst reason s to mprove utlzaton. The proposed approach provdes a mechansm for programmers to effcently utlze the dle resources located n cloud systems. The prelmnary results presented n ths study show that explotng fne gran parallelsm s promsng. Second, the dffcultes resultng from system heterogenety and the message-passng communcaton can be overcome by advanced technques, whch also motvate novel research topcs. Therefore, a number of researches focus on explotng fne gran parallelsm for loop schedulng and data mnng n cloud systems, such as [25-29]. In Secton 3.1, we menton that there are two knds of attrbutes assocated wth nodes, constants and varables. It s an nterestng ssue to nvestgate the relatonshp between these two knds of attrbutes. We thnk that each devce n a cloud system can be assocated wth these two knds of attrbutes. Takng CPU for example, CPU clock speed s a constant attrbute whle CPU loadng s a varable attrbute. Wth respect to the relatonshp between the two knds, t s ntutve that the node wth hgh CPU speed wll get more tasks to execute, resultng n hgh CPU loadng. It s probable that other devces also reveal smlar propertes. However, ths work does not focus on ths topc. We plan to take ths relatonshp nto further consderaton n our future work. In ths work, we prmarly propose a useful cloud programmng pattern, whch adopts a performance-based heurstc to dstrbute workloads, for master-slave applcatons. However, we beleve that t s possble to extend ths approach to non-master-slave applcatons, such as P2P applcatons. We explan the reason as follows. The programmng pattern abstracts our experences n programmng master-slave V. CONCLUSIONS We have proposed a programmng pattern for programmers to easly develop hgh performance applcatons on dynamc and heterogeneous cloud envronments. Ths pattern uses a performance-based approach to dstrbute workloads wthn a program to workng nodes. In ths approach, the system heterogenety s estmated by performance functons, and the varaton of workload s estmated by Statc-Workload Rato. On our cloud platform, programs mplemented by the proposed approach can obtan performance mprovement on prevous schemes. In the near future, we wll mplement more types of applcaton programs to verfy our approach. Also, applcaton of performancebased programmng to e-learnng wll be nvestgated. ACKNOWLEDGMENT The work s supported n part by Natonal Scence Councl, Tawan R.O.C., under grant no. NSC S , NSC S , NSC S MY3 and NSC S MY3. REFERENCES [1] Foster, I., The Grd: A New Infrastructure for 21st Century Scence. Physcs Today, (2): p [2] Foster, I. and C. Kesselman, Globus: A Metacomputng Infrastructure Toolkt. Internatonal Journal of Supercomputer Applcatons and Hgh Performance Computng, (2): p [3] Boeres, C., et al., An EasyGrd portal for schedulng system-aware applcatons on computatonal Grds. Concurrency and Computaton: Practce and Experence, (6): p [4] Boeres, C. and V.E.F. Rebello, EasyGrd: towards a framework for the automatc Grd enablng of legacy MPI applcatons. Concurrency and Computaton: Practce and Experence, (5): p [5] Nascmento, A.P., et al., Dstrbuted and dynamc self-schedulng of parallel MPI Grd applcatons. Concurrency and Computaton: Practce and Experence, n press. [6] Beaumont, O., et al., Schedulng dvsble loads on star and tree networks: results and open problems. Parallel and Dstrbuted Systems, IEEE Transactons on, (3): p [7] Macej, D. and L. Marcn, Mult-nstallment Dvsble Load Processng n Heterogeneous Systems wth Lmted Memory. : Parallel Processng and Appled Mathematcs [8] Macej, D. and L. Marcn, On Optmum Mult-nstallment Dvsble Load Processng n Heterogeneous Dstrbuted Systems. : Euro-Par 2005 Parallel Processng [9] Kruskal, C. and A. Wess, Allocatng ndependent subtaskson parallel processors. IEEE Transactons on Software Engneerng, : p

8 [10]Polychronopoulos, C.D. and D.J. Kuck, Guded self-schedulng: A practcal schedulng scheme for parallel supercomputers. 1987, IEEE Computer Socety. p [11]Susan Flynn, H., S. Edth, and E.F. Lawrence, Factorng: a method for schedulng parallel loops. 1992, ACM Press. p [12]Tzen, T.H. and L.M. N, Trapezod Self-Schedulng: A Practcal Schedulng Scheme for Parallel Complers. IEEE Transactons on Parallel and Dstrbuted Systems, (1): p [13]Yang, C.T. and S.C. Chang, A Parallel Loop Self-Schedulng on Extremely Heterogeneous PC Clusters. Journal of Informaton Scence and Engneerng, (2): p [14]Yang, C.T., K.W. Cheng, and K.C. L, An Effcent Parallel Loop Self- Schedulng Scheme for Cluster Envronments. Journal of Supercomputng, : p [15]Shh, W.-C., C.-T. Yang, and S.-S. Tseng, A performance-based parallel loop schedulng on grd envronments. The Journal of Supercomputng, (3): p [16]Yang, C.-T., W.-C. Shh, and S.-S. Tseng, Dynamc Parttonng of Loop Iteratons on Heterogeneous PC Clusters. to appear n The Journal of Supercomputng, [17]THU. The TIGER Grd [cted 2006; Avalable from: [18]Globus. The Globus Project [cted 2006; Avalable from: [19]MPICH. MPICH-G [cted 2006; Avalable from: [20]Han, J. and M. Kamber, Data Mnng: Concepts and Technques. 2001: Morgan Kaufmann. [21]Agrawal, R. and J.C. Shafer, Parallel Mnng of Assocaton Rules. IEEE Transactons on Knowledge and Data Engneerng, (6): p [22]Agrawal, R. and R. Srkant. Fast algorthms for Mnng Assocaton Rules. n Proc. 20th Very Large Data Bases Conf [23]Zak, M.J., Parallel and Dstrbuted Assocaton Mnng: A Survey. IEEE Concurrency, (4): p [24]Mandelbrot, B.B., Fractal Geometry of Nature. 1988, New York: W. H. Freeman. [25]Herrera, J., et al. Loosely-coupled loop schedulng n computatonal grds. n The 20th Internatonal Parallel and Dstrbuted Processng Symposum (IPDPS 2006) [26]Penmatsa, S., et al. Implementaton of Dstrbuted Loop Schedulng Schemes on the TeraGrd. n IEEE Internatonal Parallel and Dstrbuted Processng Symposum (IPDPS 2007) [27]Cannataro, M., et al., Dstrbuted data mnng on grds: servces, tools, and applcatons. Systems, Man, and Cybernetcs, Part B, IEEE Transactons on, (6): p [28]Folet, V., et al. Optmal Grd Explotaton Algorthms for Data Mnng. n The Ffth Internatonal Symposum on Parallel and Dstrbuted Computng (ISPDC '06) [29]Jang, W.-S. and J.-H. Yu. Dstrbuted data mnng on the grd. n Proceedngs of 2005 Internatonal Conference on Machne Learnng and Cybernetcs [30]Shh, W.C., Yang C.T. and Tseng S.S., Usng a Performance-based Skeleton to Implement Dvsble Load Applcatons on Grd Computng Envronments, Journal of Informaton Scence and Engneerng (JISE), 25(1), pp , [31]Apache, Hadoop, [32]J. Dean and S. Ghemawat, "Mapreduce: Smplfed data processng on large clusters," n OSDI'04: Proceedngs of the 6th Symposum on Operatng System Desgn and Implementaton. San Francsco, Calforna, USA. USENIX Assocaton, pp December 6-8, [33]THU, Cloud Testbed,

Parallelism for Nested Loops with Non-uniform and Flow Dependences

Parallelism for Nested Loops with Non-uniform and Flow Dependences Parallelsm for Nested Loops wth Non-unform and Flow Dependences Sam-Jn Jeong Dept. of Informaton & Communcaton Engneerng, Cheonan Unversty, 5, Anseo-dong, Cheonan, Chungnam, 330-80, Korea. seong@cheonan.ac.kr

More information

Concurrent Apriori Data Mining Algorithms

Concurrent Apriori Data Mining Algorithms Concurrent Apror Data Mnng Algorthms Vassl Halatchev Department of Electrcal Engneerng and Computer Scence York Unversty, Toronto October 8, 2015 Outlne Why t s mportant Introducton to Assocaton Rule Mnng

More information

Cluster Analysis of Electrical Behavior

Cluster Analysis of Electrical Behavior Journal of Computer and Communcatons, 205, 3, 88-93 Publshed Onlne May 205 n ScRes. http://www.scrp.org/ournal/cc http://dx.do.org/0.4236/cc.205.350 Cluster Analyss of Electrcal Behavor Ln Lu Ln Lu, School

More information

Comparison of Heuristics for Scheduling Independent Tasks on Heterogeneous Distributed Environments

Comparison of Heuristics for Scheduling Independent Tasks on Heterogeneous Distributed Environments Comparson of Heurstcs for Schedulng Independent Tasks on Heterogeneous Dstrbuted Envronments Hesam Izakan¹, Ath Abraham², Senor Member, IEEE, Václav Snášel³ ¹ Islamc Azad Unversty, Ramsar Branch, Ramsar,

More information

Improvement of Spatial Resolution Using BlockMatching Based Motion Estimation and Frame. Integration

Improvement of Spatial Resolution Using BlockMatching Based Motion Estimation and Frame. Integration Improvement of Spatal Resoluton Usng BlockMatchng Based Moton Estmaton and Frame Integraton Danya Suga and Takayuk Hamamoto Graduate School of Engneerng, Tokyo Unversty of Scence, 6-3-1, Nuku, Katsuska-ku,

More information

Distributed Resource Scheduling in Grid Computing Using Fuzzy Approach

Distributed Resource Scheduling in Grid Computing Using Fuzzy Approach Dstrbuted Resource Schedulng n Grd Computng Usng Fuzzy Approach Shahram Amn, Mohammad Ahmad Computer Engneerng Department Islamc Azad Unversty branch Mahallat, Iran Islamc Azad Unversty branch khomen,

More information

NUMERICAL SOLVING OPTIMAL CONTROL PROBLEMS BY THE METHOD OF VARIATIONS

NUMERICAL SOLVING OPTIMAL CONTROL PROBLEMS BY THE METHOD OF VARIATIONS ARPN Journal of Engneerng and Appled Scences 006-017 Asan Research Publshng Network (ARPN). All rghts reserved. NUMERICAL SOLVING OPTIMAL CONTROL PROBLEMS BY THE METHOD OF VARIATIONS Igor Grgoryev, Svetlana

More information

Parallel matrix-vector multiplication

Parallel matrix-vector multiplication Appendx A Parallel matrx-vector multplcaton The reduced transton matrx of the three-dmensonal cage model for gel electrophoress, descrbed n secton 3.2, becomes excessvely large for polymer lengths more

More information

An Iterative Solution Approach to Process Plant Layout using Mixed Integer Optimisation

An Iterative Solution Approach to Process Plant Layout using Mixed Integer Optimisation 17 th European Symposum on Computer Aded Process Engneerng ESCAPE17 V. Plesu and P.S. Agach (Edtors) 2007 Elsever B.V. All rghts reserved. 1 An Iteratve Soluton Approach to Process Plant Layout usng Mxed

More information

An Application of the Dulmage-Mendelsohn Decomposition to Sparse Null Space Bases of Full Row Rank Matrices

An Application of the Dulmage-Mendelsohn Decomposition to Sparse Null Space Bases of Full Row Rank Matrices Internatonal Mathematcal Forum, Vol 7, 2012, no 52, 2549-2554 An Applcaton of the Dulmage-Mendelsohn Decomposton to Sparse Null Space Bases of Full Row Rank Matrces Mostafa Khorramzadeh Department of Mathematcal

More information

A Binarization Algorithm specialized on Document Images and Photos

A Binarization Algorithm specialized on Document Images and Photos A Bnarzaton Algorthm specalzed on Document mages and Photos Ergna Kavalleratou Dept. of nformaton and Communcaton Systems Engneerng Unversty of the Aegean kavalleratou@aegean.gr Abstract n ths paper, a

More information

Classifier Selection Based on Data Complexity Measures *

Classifier Selection Based on Data Complexity Measures * Classfer Selecton Based on Data Complexty Measures * Edth Hernández-Reyes, J.A. Carrasco-Ochoa, and J.Fco. Martínez-Trndad Natonal Insttute for Astrophyscs, Optcs and Electroncs, Lus Enrque Erro No.1 Sta.

More information

Simulation Based Analysis of FAST TCP using OMNET++

Simulation Based Analysis of FAST TCP using OMNET++ Smulaton Based Analyss of FAST TCP usng OMNET++ Umar ul Hassan 04030038@lums.edu.pk Md Term Report CS678 Topcs n Internet Research Sprng, 2006 Introducton Internet traffc s doublng roughly every 3 months

More information

Scheduling Remote Access to Scientific Instruments in Cyberinfrastructure for Education and Research

Scheduling Remote Access to Scientific Instruments in Cyberinfrastructure for Education and Research Schedulng Remote Access to Scentfc Instruments n Cybernfrastructure for Educaton and Research Je Yn 1, Junwe Cao 2,3,*, Yuexuan Wang 4, Lanchen Lu 1,3 and Cheng Wu 1,3 1 Natonal CIMS Engneerng and Research

More information

Support Vector Machines

Support Vector Machines /9/207 MIST.6060 Busness Intellgence and Data Mnng What are Support Vector Machnes? Support Vector Machnes Support Vector Machnes (SVMs) are supervsed learnng technques that analyze data and recognze patterns.

More information

An Optimal Algorithm for Prufer Codes *

An Optimal Algorithm for Prufer Codes * J. Software Engneerng & Applcatons, 2009, 2: 111-115 do:10.4236/jsea.2009.22016 Publshed Onlne July 2009 (www.scrp.org/journal/jsea) An Optmal Algorthm for Prufer Codes * Xaodong Wang 1, 2, Le Wang 3,

More information

TECHNIQUE OF FORMATION HOMOGENEOUS SAMPLE SAME OBJECTS. Muradaliyev A.Z.

TECHNIQUE OF FORMATION HOMOGENEOUS SAMPLE SAME OBJECTS. Muradaliyev A.Z. TECHNIQUE OF FORMATION HOMOGENEOUS SAMPLE SAME OBJECTS Muradalyev AZ Azerbajan Scentfc-Research and Desgn-Prospectng Insttute of Energetc AZ1012, Ave HZardab-94 E-mal:aydn_murad@yahoocom Importance of

More information

Load Balancing for Hex-Cell Interconnection Network

Load Balancing for Hex-Cell Interconnection Network Int. J. Communcatons, Network and System Scences,,, - Publshed Onlne Aprl n ScRes. http://www.scrp.org/journal/jcns http://dx.do.org/./jcns.. Load Balancng for Hex-Cell Interconnecton Network Saher Manaseer,

More information

Compiler Design. Spring Register Allocation. Sample Exercises and Solutions. Prof. Pedro C. Diniz

Compiler Design. Spring Register Allocation. Sample Exercises and Solutions. Prof. Pedro C. Diniz Compler Desgn Sprng 2014 Regster Allocaton Sample Exercses and Solutons Prof. Pedro C. Dnz USC / Informaton Scences Insttute 4676 Admralty Way, Sute 1001 Marna del Rey, Calforna 90292 pedro@s.edu Regster

More information

A Fast Visual Tracking Algorithm Based on Circle Pixels Matching

A Fast Visual Tracking Algorithm Based on Circle Pixels Matching A Fast Vsual Trackng Algorthm Based on Crcle Pxels Matchng Zhqang Hou hou_zhq@sohu.com Chongzhao Han czhan@mal.xjtu.edu.cn Ln Zheng Abstract: A fast vsual trackng algorthm based on crcle pxels matchng

More information

Parallel and Distributed Association Rule Mining - Dr. Giuseppe Di Fatta. San Vigilio,

Parallel and Distributed Association Rule Mining - Dr. Giuseppe Di Fatta. San Vigilio, Parallel and Dstrbuted Assocaton Rule Mnng - Dr. Guseppe D Fatta fatta@nf.un-konstanz.de San Vglo, 18-09-2004 1 Overvew Assocaton Rule Mnng (ARM) Apror algorthm Hgh Performance Parallel and Dstrbuted Computng

More information

Subspace clustering. Clustering. Fundamental to all clustering techniques is the choice of distance measure between data points;

Subspace clustering. Clustering. Fundamental to all clustering techniques is the choice of distance measure between data points; Subspace clusterng Clusterng Fundamental to all clusterng technques s the choce of dstance measure between data ponts; D q ( ) ( ) 2 x x = x x, j k = 1 k jk Squared Eucldean dstance Assumpton: All features

More information

Video Proxy System for a Large-scale VOD System (DINA)

Video Proxy System for a Large-scale VOD System (DINA) Vdeo Proxy System for a Large-scale VOD System (DINA) KWUN-CHUNG CHAN #, KWOK-WAI CHEUNG *# #Department of Informaton Engneerng *Centre of Innovaton and Technology The Chnese Unversty of Hong Kong SHATIN,

More information

Some material adapted from Mohamed Younis, UMBC CMSC 611 Spr 2003 course slides Some material adapted from Hennessy & Patterson / 2003 Elsevier

Some material adapted from Mohamed Younis, UMBC CMSC 611 Spr 2003 course slides Some material adapted from Hennessy & Patterson / 2003 Elsevier Some materal adapted from Mohamed Youns, UMBC CMSC 611 Spr 2003 course sldes Some materal adapted from Hennessy & Patterson / 2003 Elsever Scence Performance = 1 Executon tme Speedup = Performance (B)

More information

The Research of Support Vector Machine in Agricultural Data Classification

The Research of Support Vector Machine in Agricultural Data Classification The Research of Support Vector Machne n Agrcultural Data Classfcaton Le Sh, Qguo Duan, Xnmng Ma, Me Weng College of Informaton and Management Scence, HeNan Agrcultural Unversty, Zhengzhou 45000 Chna Zhengzhou

More information

Learning-Based Top-N Selection Query Evaluation over Relational Databases

Learning-Based Top-N Selection Query Evaluation over Relational Databases Learnng-Based Top-N Selecton Query Evaluaton over Relatonal Databases Lang Zhu *, Wey Meng ** * School of Mathematcs and Computer Scence, Hebe Unversty, Baodng, Hebe 071002, Chna, zhu@mal.hbu.edu.cn **

More information

Helsinki University Of Technology, Systems Analysis Laboratory Mat Independent research projects in applied mathematics (3 cr)

Helsinki University Of Technology, Systems Analysis Laboratory Mat Independent research projects in applied mathematics (3 cr) Helsnk Unversty Of Technology, Systems Analyss Laboratory Mat-2.08 Independent research projects n appled mathematcs (3 cr) "! #$&% Antt Laukkanen 506 R ajlaukka@cc.hut.f 2 Introducton...3 2 Multattrbute

More information

Determining the Optimal Bandwidth Based on Multi-criterion Fusion

Determining the Optimal Bandwidth Based on Multi-criterion Fusion Proceedngs of 01 4th Internatonal Conference on Machne Learnng and Computng IPCSIT vol. 5 (01) (01) IACSIT Press, Sngapore Determnng the Optmal Bandwdth Based on Mult-crteron Fuson Ha-L Lang 1+, Xan-Mn

More information

Two-Stage Data Distribution for Distributed Surveillance Video Processing with Hybrid Storage Architecture

Two-Stage Data Distribution for Distributed Surveillance Video Processing with Hybrid Storage Architecture Two-Stage Data Dstrbuton for Dstrbuted Survellance Vdeo Processng wth Hybrd Storage Archtecture Yangyang Gao, Hatao Zhang, Bngchang Tang, Yanpe Zhu, Huadong Ma Bejng Key Lab of Intellgent Telecomm. Software

More information

SLAM Summer School 2006 Practical 2: SLAM using Monocular Vision

SLAM Summer School 2006 Practical 2: SLAM using Monocular Vision SLAM Summer School 2006 Practcal 2: SLAM usng Monocular Vson Javer Cvera, Unversty of Zaragoza Andrew J. Davson, Imperal College London J.M.M Montel, Unversty of Zaragoza. josemar@unzar.es, jcvera@unzar.es,

More information

Real-time Fault-tolerant Scheduling Algorithm for Distributed Computing Systems

Real-time Fault-tolerant Scheduling Algorithm for Distributed Computing Systems Real-tme Fault-tolerant Schedulng Algorthm for Dstrbuted Computng Systems Yun Lng, Y Ouyang College of Computer Scence and Informaton Engneerng Zheang Gongshang Unversty Postal code: 310018 P.R.CHINA {ylng,

More information

The Codesign Challenge

The Codesign Challenge ECE 4530 Codesgn Challenge Fall 2007 Hardware/Software Codesgn The Codesgn Challenge Objectves In the codesgn challenge, your task s to accelerate a gven software reference mplementaton as fast as possble.

More information

Term Weighting Classification System Using the Chi-square Statistic for the Classification Subtask at NTCIR-6 Patent Retrieval Task

Term Weighting Classification System Using the Chi-square Statistic for the Classification Subtask at NTCIR-6 Patent Retrieval Task Proceedngs of NTCIR-6 Workshop Meetng, May 15-18, 2007, Tokyo, Japan Term Weghtng Classfcaton System Usng the Ch-square Statstc for the Classfcaton Subtask at NTCIR-6 Patent Retreval Task Kotaro Hashmoto

More information

Load-Balanced Anycast Routing

Load-Balanced Anycast Routing Load-Balanced Anycast Routng Chng-Yu Ln, Jung-Hua Lo, and Sy-Yen Kuo Department of Electrcal Engneerng atonal Tawan Unversty, Tape, Tawan sykuo@cc.ee.ntu.edu.tw Abstract For fault-tolerance and load-balance

More information

Query Clustering Using a Hybrid Query Similarity Measure

Query Clustering Using a Hybrid Query Similarity Measure Query clusterng usng a hybrd query smlarty measure Fu. L., Goh, D.H., & Foo, S. (2004). WSEAS Transacton on Computers, 3(3), 700-705. Query Clusterng Usng a Hybrd Query Smlarty Measure Ln Fu, Don Hoe-Lan

More information

Classifying Acoustic Transient Signals Using Artificial Intelligence

Classifying Acoustic Transient Signals Using Artificial Intelligence Classfyng Acoustc Transent Sgnals Usng Artfcal Intellgence Steve Sutton, Unversty of North Carolna At Wlmngton (suttons@charter.net) Greg Huff, Unversty of North Carolna At Wlmngton (jgh7476@uncwl.edu)

More information

A mathematical programming approach to the analysis, design and scheduling of offshore oilfields

A mathematical programming approach to the analysis, design and scheduling of offshore oilfields 17 th European Symposum on Computer Aded Process Engneerng ESCAPE17 V. Plesu and P.S. Agach (Edtors) 2007 Elsever B.V. All rghts reserved. 1 A mathematcal programmng approach to the analyss, desgn and

More information

Wishing you all a Total Quality New Year!

Wishing you all a Total Quality New Year! Total Qualty Management and Sx Sgma Post Graduate Program 214-15 Sesson 4 Vnay Kumar Kalakband Assstant Professor Operatons & Systems Area 1 Wshng you all a Total Qualty New Year! Hope you acheve Sx sgma

More information

Learning the Kernel Parameters in Kernel Minimum Distance Classifier

Learning the Kernel Parameters in Kernel Minimum Distance Classifier Learnng the Kernel Parameters n Kernel Mnmum Dstance Classfer Daoqang Zhang 1,, Songcan Chen and Zh-Hua Zhou 1* 1 Natonal Laboratory for Novel Software Technology Nanjng Unversty, Nanjng 193, Chna Department

More information

Content Based Image Retrieval Using 2-D Discrete Wavelet with Texture Feature with Different Classifiers

Content Based Image Retrieval Using 2-D Discrete Wavelet with Texture Feature with Different Classifiers IOSR Journal of Electroncs and Communcaton Engneerng (IOSR-JECE) e-issn: 78-834,p- ISSN: 78-8735.Volume 9, Issue, Ver. IV (Mar - Apr. 04), PP 0-07 Content Based Image Retreval Usng -D Dscrete Wavelet wth

More information

A MOVING MESH APPROACH FOR SIMULATION BUDGET ALLOCATION ON CONTINUOUS DOMAINS

A MOVING MESH APPROACH FOR SIMULATION BUDGET ALLOCATION ON CONTINUOUS DOMAINS Proceedngs of the Wnter Smulaton Conference M E Kuhl, N M Steger, F B Armstrong, and J A Jones, eds A MOVING MESH APPROACH FOR SIMULATION BUDGET ALLOCATION ON CONTINUOUS DOMAINS Mark W Brantley Chun-Hung

More information

A Fast Content-Based Multimedia Retrieval Technique Using Compressed Data

A Fast Content-Based Multimedia Retrieval Technique Using Compressed Data A Fast Content-Based Multmeda Retreval Technque Usng Compressed Data Borko Furht and Pornvt Saksobhavvat NSF Multmeda Laboratory Florda Atlantc Unversty, Boca Raton, Florda 3343 ABSTRACT In ths paper,

More information

Virtual Memory. Background. No. 10. Virtual Memory: concept. Logical Memory Space (review) Demand Paging(1) Virtual Memory

Virtual Memory. Background. No. 10. Virtual Memory: concept. Logical Memory Space (review) Demand Paging(1) Virtual Memory Background EECS. Operatng System Fundamentals No. Vrtual Memory Prof. Hu Jang Department of Electrcal Engneerng and Computer Scence, York Unversty Memory-management methods normally requres the entre process

More information

Hybrid Job Scheduling Mechanism Using a Backfill-based Multi-queue Strategy in Distributed Grid Computing

Hybrid Job Scheduling Mechanism Using a Backfill-based Multi-queue Strategy in Distributed Grid Computing IJCSNS Internatonal Journal of Computer Scence and Network Securty, VOL.12 No.9, September 2012 39 Hybrd Job Schedulng Mechansm Usng a Backfll-based Mult-queue Strategy n Dstrbuted Grd Computng Ken Park

More information

Virtual Machine Migration based on Trust Measurement of Computer Node

Virtual Machine Migration based on Trust Measurement of Computer Node Appled Mechancs and Materals Onlne: 2014-04-04 ISSN: 1662-7482, Vols. 536-537, pp 678-682 do:10.4028/www.scentfc.net/amm.536-537.678 2014 Trans Tech Publcatons, Swtzerland Vrtual Machne Mgraton based on

More information

Course Introduction. Algorithm 8/31/2017. COSC 320 Advanced Data Structures and Algorithms. COSC 320 Advanced Data Structures and Algorithms

Course Introduction. Algorithm 8/31/2017. COSC 320 Advanced Data Structures and Algorithms. COSC 320 Advanced Data Structures and Algorithms Course Introducton Course Topcs Exams, abs, Proects A quc loo at a few algorthms 1 Advanced Data Structures and Algorthms Descrpton: We are gong to dscuss algorthm complexty analyss, algorthm desgn technques

More information

Solving two-person zero-sum game by Matlab

Solving two-person zero-sum game by Matlab Appled Mechancs and Materals Onlne: 2011-02-02 ISSN: 1662-7482, Vols. 50-51, pp 262-265 do:10.4028/www.scentfc.net/amm.50-51.262 2011 Trans Tech Publcatons, Swtzerland Solvng two-person zero-sum game by

More information

X- Chart Using ANOM Approach

X- Chart Using ANOM Approach ISSN 1684-8403 Journal of Statstcs Volume 17, 010, pp. 3-3 Abstract X- Chart Usng ANOM Approach Gullapall Chakravarth 1 and Chaluvad Venkateswara Rao Control lmts for ndvdual measurements (X) chart are

More information

A Model Based on Multi-agent for Dynamic Bandwidth Allocation in Networks Guang LU, Jian-Wen QI

A Model Based on Multi-agent for Dynamic Bandwidth Allocation in Networks Guang LU, Jian-Wen QI 216 Jont Internatonal Conference on Artfcal Intellgence and Computer Engneerng (AICE 216) and Internatonal Conference on etwork and Communcaton Securty (CS 216) ISB: 978-1-6595-362-5 A Model Based on Mult-agent

More information

CHAPTER 2 PROPOSED IMPROVED PARTICLE SWARM OPTIMIZATION

CHAPTER 2 PROPOSED IMPROVED PARTICLE SWARM OPTIMIZATION 24 CHAPTER 2 PROPOSED IMPROVED PARTICLE SWARM OPTIMIZATION The present chapter proposes an IPSO approach for multprocessor task schedulng problem wth two classfcatons, namely, statc ndependent tasks and

More information

An Entropy-Based Approach to Integrated Information Needs Assessment

An Entropy-Based Approach to Integrated Information Needs Assessment Dstrbuton Statement A: Approved for publc release; dstrbuton s unlmted. An Entropy-Based Approach to ntegrated nformaton Needs Assessment June 8, 2004 Wllam J. Farrell Lockheed Martn Advanced Technology

More information

Configuration Management in Multi-Context Reconfigurable Systems for Simultaneous Performance and Power Optimizations*

Configuration Management in Multi-Context Reconfigurable Systems for Simultaneous Performance and Power Optimizations* Confguraton Management n Mult-Context Reconfgurable Systems for Smultaneous Performance and Power Optmzatons* Rafael Maestre, Mlagros Fernandez Departamento de Arqutectura de Computadores y Automátca Unversdad

More information

Meta-heuristics for Multidimensional Knapsack Problems

Meta-heuristics for Multidimensional Knapsack Problems 2012 4th Internatonal Conference on Computer Research and Development IPCSIT vol.39 (2012) (2012) IACSIT Press, Sngapore Meta-heurstcs for Multdmensonal Knapsack Problems Zhbao Man + Computer Scence Department,

More information

A Background Subtraction for a Vision-based User Interface *

A Background Subtraction for a Vision-based User Interface * A Background Subtracton for a Vson-based User Interface * Dongpyo Hong and Woontack Woo KJIST U-VR Lab. {dhon wwoo}@kjst.ac.kr Abstract In ths paper, we propose a robust and effcent background subtracton

More information

UB at GeoCLEF Department of Geography Abstract

UB at GeoCLEF Department of Geography   Abstract UB at GeoCLEF 2006 Mguel E. Ruz (1), Stuart Shapro (2), June Abbas (1), Slva B. Southwck (1) and Davd Mark (3) State Unversty of New York at Buffalo (1) Department of Lbrary and Informaton Studes (2) Department

More information

Fuzzy Filtering Algorithms for Image Processing: Performance Evaluation of Various Approaches

Fuzzy Filtering Algorithms for Image Processing: Performance Evaluation of Various Approaches Proceedngs of the Internatonal Conference on Cognton and Recognton Fuzzy Flterng Algorthms for Image Processng: Performance Evaluaton of Varous Approaches Rajoo Pandey and Umesh Ghanekar Department of

More information

Programming in Fortran 90 : 2017/2018

Programming in Fortran 90 : 2017/2018 Programmng n Fortran 90 : 2017/2018 Programmng n Fortran 90 : 2017/2018 Exercse 1 : Evaluaton of functon dependng on nput Wrte a program who evaluate the functon f (x,y) for any two user specfed values

More information

Study of Data Stream Clustering Based on Bio-inspired Model

Study of Data Stream Clustering Based on Bio-inspired Model , pp.412-418 http://dx.do.org/10.14257/astl.2014.53.86 Study of Data Stream lusterng Based on Bo-nspred Model Yngme L, Mn L, Jngbo Shao, Gaoyang Wang ollege of omputer Scence and Informaton Engneerng,

More information

Scheduling Independent Tasks in Heterogeneous Environments under Communication Constraints

Scheduling Independent Tasks in Heterogeneous Environments under Communication Constraints Schedulng Independent Tasks n Heterogeneous Envronments under Communcaton Constrants Petros Lampsas 1 Thanass Loukopoulos 2 Fedon Dmopoulos 1 Marouso Athanasou 1 1 Department of Informatcs and Computer

More information

A Genetic Algorithm Based Dynamic Load Balancing Scheme for Heterogeneous Distributed Systems

A Genetic Algorithm Based Dynamic Load Balancing Scheme for Heterogeneous Distributed Systems Proceedngs of the Internatonal Conference on Parallel and Dstrbuted Processng Technques and Applcatons, PDPTA 2008, Las Vegas, Nevada, USA, July 14-17, 2008, 2 Volumes. CSREA Press 2008, ISBN 1-60132-084-1

More information

Performance Evaluation of Information Retrieval Systems

Performance Evaluation of Information Retrieval Systems Why System Evaluaton? Performance Evaluaton of Informaton Retreval Systems Many sldes n ths secton are adapted from Prof. Joydeep Ghosh (UT ECE) who n turn adapted them from Prof. Dk Lee (Unv. of Scence

More information

Cache Performance 3/28/17. Agenda. Cache Abstraction and Metrics. Direct-Mapped Cache: Placement and Access

Cache Performance 3/28/17. Agenda. Cache Abstraction and Metrics. Direct-Mapped Cache: Placement and Access Agenda Cache Performance Samra Khan March 28, 217 Revew from last lecture Cache access Assocatvty Replacement Cache Performance Cache Abstracton and Metrcs Address Tag Store (s the address n the cache?

More information

THE PULL-PUSH ALGORITHM REVISITED

THE PULL-PUSH ALGORITHM REVISITED THE PULL-PUSH ALGORITHM REVISITED Improvements, Computaton of Pont Denstes, and GPU Implementaton Martn Kraus Computer Graphcs & Vsualzaton Group, Technsche Unverstät München, Boltzmannstraße 3, 85748

More information

FINDING IMPORTANT NODES IN SOCIAL NETWORKS BASED ON MODIFIED PAGERANK

FINDING IMPORTANT NODES IN SOCIAL NETWORKS BASED ON MODIFIED PAGERANK FINDING IMPORTANT NODES IN SOCIAL NETWORKS BASED ON MODIFIED PAGERANK L-qng Qu, Yong-quan Lang 2, Jng-Chen 3, 2 College of Informaton Scence and Technology, Shandong Unversty of Scence and Technology,

More information

Application of VCG in Replica Placement Strategy of Cloud Storage

Application of VCG in Replica Placement Strategy of Cloud Storage Internatonal Journal of Grd and Dstrbuted Computng, pp.27-40 http://dx.do.org/10.14257/jgdc.2016.9.4.03 Applcaton of VCG n Replca Placement Strategy of Cloud Storage Wang Hongxa Computer Department, Bejng

More information

Problem Definitions and Evaluation Criteria for Computational Expensive Optimization

Problem Definitions and Evaluation Criteria for Computational Expensive Optimization Problem efntons and Evaluaton Crtera for Computatonal Expensve Optmzaton B. Lu 1, Q. Chen and Q. Zhang 3, J. J. Lang 4, P. N. Suganthan, B. Y. Qu 6 1 epartment of Computng, Glyndwr Unversty, UK Faclty

More information

Reducing Frame Rate for Object Tracking

Reducing Frame Rate for Object Tracking Reducng Frame Rate for Object Trackng Pavel Korshunov 1 and We Tsang Oo 2 1 Natonal Unversty of Sngapore, Sngapore 11977, pavelkor@comp.nus.edu.sg 2 Natonal Unversty of Sngapore, Sngapore 11977, oowt@comp.nus.edu.sg

More information

BOOSTING CLASSIFICATION ACCURACY WITH SAMPLES CHOSEN FROM A VALIDATION SET

BOOSTING CLASSIFICATION ACCURACY WITH SAMPLES CHOSEN FROM A VALIDATION SET 1 BOOSTING CLASSIFICATION ACCURACY WITH SAMPLES CHOSEN FROM A VALIDATION SET TZU-CHENG CHUANG School of Electrcal and Computer Engneerng, Purdue Unversty, West Lafayette, Indana 47907 SAUL B. GELFAND School

More information

AADL : about scheduling analysis

AADL : about scheduling analysis AADL : about schedulng analyss Schedulng analyss, what s t? Embedded real-tme crtcal systems have temporal constrants to meet (e.g. deadlne). Many systems are bult wth operatng systems provdng multtaskng

More information

Improved Resource Allocation Algorithms for Practical Image Encoding in a Ubiquitous Computing Environment

Improved Resource Allocation Algorithms for Practical Image Encoding in a Ubiquitous Computing Environment JOURNAL OF COMPUTERS, VOL. 4, NO. 9, SEPTEMBER 2009 873 Improved Resource Allocaton Algorthms for Practcal Image Encodng n a Ubqutous Computng Envronment Manxong Dong, Long Zheng, Kaoru Ota, Song Guo School

More information

Module Management Tool in Software Development Organizations

Module Management Tool in Software Development Organizations Journal of Computer Scence (5): 8-, 7 ISSN 59-66 7 Scence Publcatons Management Tool n Software Development Organzatons Ahmad A. Al-Rababah and Mohammad A. Al-Rababah Faculty of IT, Al-Ahlyyah Amman Unversty,

More information

EVALUATION OF THE PERFORMANCES OF ARTIFICIAL BEE COLONY AND INVASIVE WEED OPTIMIZATION ALGORITHMS ON THE MODIFIED BENCHMARK FUNCTIONS

EVALUATION OF THE PERFORMANCES OF ARTIFICIAL BEE COLONY AND INVASIVE WEED OPTIMIZATION ALGORITHMS ON THE MODIFIED BENCHMARK FUNCTIONS Academc Research Internatonal ISS-L: 3-9553, ISS: 3-9944 Vol., o. 3, May 0 EVALUATIO OF THE PERFORMACES OF ARTIFICIAL BEE COLOY AD IVASIVE WEED OPTIMIZATIO ALGORITHMS O THE MODIFIED BECHMARK FUCTIOS Dlay

More information

Incremental Learning with Support Vector Machines and Fuzzy Set Theory

Incremental Learning with Support Vector Machines and Fuzzy Set Theory The 25th Workshop on Combnatoral Mathematcs and Computaton Theory Incremental Learnng wth Support Vector Machnes and Fuzzy Set Theory Yu-Mng Chuang 1 and Cha-Hwa Ln 2* 1 Department of Computer Scence and

More information

Intelligent Information Acquisition for Improved Clustering

Intelligent Information Acquisition for Improved Clustering Intellgent Informaton Acquston for Improved Clusterng Duy Vu Unversty of Texas at Austn duyvu@cs.utexas.edu Mkhal Blenko Mcrosoft Research mblenko@mcrosoft.com Prem Melvlle IBM T.J. Watson Research Center

More information

S1 Note. Basis functions.

S1 Note. Basis functions. S1 Note. Bass functons. Contents Types of bass functons...1 The Fourer bass...2 B-splne bass...3 Power and type I error rates wth dfferent numbers of bass functons...4 Table S1. Smulaton results of type

More information

APPLICATION OF MULTIVARIATE LOSS FUNCTION FOR ASSESSMENT OF THE QUALITY OF TECHNOLOGICAL PROCESS MANAGEMENT

APPLICATION OF MULTIVARIATE LOSS FUNCTION FOR ASSESSMENT OF THE QUALITY OF TECHNOLOGICAL PROCESS MANAGEMENT 3. - 5. 5., Brno, Czech Republc, EU APPLICATION OF MULTIVARIATE LOSS FUNCTION FOR ASSESSMENT OF THE QUALITY OF TECHNOLOGICAL PROCESS MANAGEMENT Abstract Josef TOŠENOVSKÝ ) Lenka MONSPORTOVÁ ) Flp TOŠENOVSKÝ

More information

Dynamic Bandwidth Provisioning with Fairness and Revenue Considerations for Broadband Wireless Communication

Dynamic Bandwidth Provisioning with Fairness and Revenue Considerations for Broadband Wireless Communication Ths full text paper was peer revewed at the drecton of IEEE Communcatons Socety subject matter experts for publcaton n the ICC 008 proceedngs. Dynamc Bandwdth Provsonng wth Farness and Revenue Consderatons

More information

User Authentication Based On Behavioral Mouse Dynamics Biometrics

User Authentication Based On Behavioral Mouse Dynamics Biometrics User Authentcaton Based On Behavoral Mouse Dynamcs Bometrcs Chee-Hyung Yoon Danel Donghyun Km Department of Computer Scence Department of Computer Scence Stanford Unversty Stanford Unversty Stanford, CA

More information

Assignment # 2. Farrukh Jabeen Algorithms 510 Assignment #2 Due Date: June 15, 2009.

Assignment # 2. Farrukh Jabeen Algorithms 510 Assignment #2 Due Date: June 15, 2009. Farrukh Jabeen Algorthms 51 Assgnment #2 Due Date: June 15, 29. Assgnment # 2 Chapter 3 Dscrete Fourer Transforms Implement the FFT for the DFT. Descrbed n sectons 3.1 and 3.2. Delverables: 1. Concse descrpton

More information

Remote Sensing Image Retrieval Algorithm based on MapReduce and Characteristic Information

Remote Sensing Image Retrieval Algorithm based on MapReduce and Characteristic Information Remote Sensng Image Retreval Algorthm based on MapReduce and Characterstc Informaton Zhang Meng 1, 1 Computer School, Wuhan Unversty Hube, Wuhan430097 Informaton Center, Wuhan Unversty Hube, Wuhan430097

More information

Hybrid Non-Blind Color Image Watermarking

Hybrid Non-Blind Color Image Watermarking Hybrd Non-Blnd Color Image Watermarkng Ms C.N.Sujatha 1, Dr. P. Satyanarayana 2 1 Assocate Professor, Dept. of ECE, SNIST, Yamnampet, Ghatkesar Hyderabad-501301, Telangana 2 Professor, Dept. of ECE, AITS,

More information

A Deflected Grid-based Algorithm for Clustering Analysis

A Deflected Grid-based Algorithm for Clustering Analysis A Deflected Grd-based Algorthm for Clusterng Analyss NANCY P. LIN, CHUNG-I CHANG, HAO-EN CHUEH, HUNG-JEN CHEN, WEI-HUA HAO Department of Computer Scence and Informaton Engneerng Tamkang Unversty 5 Yng-chuan

More information

Smoothing Spline ANOVA for variable screening

Smoothing Spline ANOVA for variable screening Smoothng Splne ANOVA for varable screenng a useful tool for metamodels tranng and mult-objectve optmzaton L. Rcco, E. Rgon, A. Turco Outlne RSM Introducton Possble couplng Test case MOO MOO wth Game Theory

More information

Efficient Distributed File System (EDFS)

Efficient Distributed File System (EDFS) Effcent Dstrbuted Fle System (EDFS) (Sem-Centralzed) Debessay(Debsh) Fesehaye, Rahul Malk & Klara Naherstedt Unversty of Illnos-Urbana Champagn Contents Problem Statement, Related Work, EDFS Desgn Rate

More information

The Shortest Path of Touring Lines given in the Plane

The Shortest Path of Touring Lines given in the Plane Send Orders for Reprnts to reprnts@benthamscence.ae 262 The Open Cybernetcs & Systemcs Journal, 2015, 9, 262-267 The Shortest Path of Tourng Lnes gven n the Plane Open Access Ljuan Wang 1,2, Dandan He

More information

Petri Net Based Software Dependability Engineering

Petri Net Based Software Dependability Engineering Proc. RELECTRONIC 95, Budapest, pp. 181-186; October 1995 Petr Net Based Software Dependablty Engneerng Monka Hener Brandenburg Unversty of Technology Cottbus Computer Scence Insttute Postbox 101344 D-03013

More information

Review of approximation techniques

Review of approximation techniques CHAPTER 2 Revew of appromaton technques 2. Introducton Optmzaton problems n engneerng desgn are characterzed by the followng assocated features: the objectve functon and constrants are mplct functons evaluated

More information

Feature Reduction and Selection

Feature Reduction and Selection Feature Reducton and Selecton Dr. Shuang LIANG School of Software Engneerng TongJ Unversty Fall, 2012 Today s Topcs Introducton Problems of Dmensonalty Feature Reducton Statstc methods Prncpal Components

More information

Maximum Variance Combined with Adaptive Genetic Algorithm for Infrared Image Segmentation

Maximum Variance Combined with Adaptive Genetic Algorithm for Infrared Image Segmentation Internatonal Conference on Logstcs Engneerng, Management and Computer Scence (LEMCS 5) Maxmum Varance Combned wth Adaptve Genetc Algorthm for Infrared Image Segmentaton Huxuan Fu College of Automaton Harbn

More information

A SYSTOLIC APPROACH TO LOOP PARTITIONING AND MAPPING INTO FIXED SIZE DISTRIBUTED MEMORY ARCHITECTURES

A SYSTOLIC APPROACH TO LOOP PARTITIONING AND MAPPING INTO FIXED SIZE DISTRIBUTED MEMORY ARCHITECTURES A SYSOLIC APPROACH O LOOP PARIIONING AND MAPPING INO FIXED SIZE DISRIBUED MEMORY ARCHIECURES Ioanns Drosts, Nektaros Kozrs, George Papakonstantnou and Panayots sanakas Natonal echncal Unversty of Athens

More information

Research of Dynamic Access to Cloud Database Based on Improved Pheromone Algorithm

Research of Dynamic Access to Cloud Database Based on Improved Pheromone Algorithm , pp.197-202 http://dx.do.org/10.14257/dta.2016.9.5.20 Research of Dynamc Access to Cloud Database Based on Improved Pheromone Algorthm Yongqang L 1 and Jn Pan 2 1 (Software Technology Vocatonal College,

More information

Array transposition in CUDA shared memory

Array transposition in CUDA shared memory Array transposton n CUDA shared memory Mke Gles February 19, 2014 Abstract Ths short note s nspred by some code wrtten by Jeremy Appleyard for the transposton of data through shared memory. I had some

More information

FEATURE EXTRACTION. Dr. K.Vijayarekha. Associate Dean School of Electrical and Electronics Engineering SASTRA University, Thanjavur

FEATURE EXTRACTION. Dr. K.Vijayarekha. Associate Dean School of Electrical and Electronics Engineering SASTRA University, Thanjavur FEATURE EXTRACTION Dr. K.Vjayarekha Assocate Dean School of Electrcal and Electroncs Engneerng SASTRA Unversty, Thanjavur613 41 Jont Intatve of IITs and IISc Funded by MHRD Page 1 of 8 Table of Contents

More information

FAHP and Modified GRA Based Network Selection in Heterogeneous Wireless Networks

FAHP and Modified GRA Based Network Selection in Heterogeneous Wireless Networks 2017 2nd Internatonal Semnar on Appled Physcs, Optoelectroncs and Photoncs (APOP 2017) ISBN: 978-1-60595-522-3 FAHP and Modfed GRA Based Network Selecton n Heterogeneous Wreless Networks Xaohan DU, Zhqng

More information

BIN XIA et al: AN IMPROVED K-MEANS ALGORITHM BASED ON CLOUD PLATFORM FOR DATA MINING

BIN XIA et al: AN IMPROVED K-MEANS ALGORITHM BASED ON CLOUD PLATFORM FOR DATA MINING An Improved K-means Algorthm based on Cloud Platform for Data Mnng Bn Xa *, Yan Lu 2. School of nformaton and management scence, Henan Agrcultural Unversty, Zhengzhou, Henan 450002, P.R. Chna 2. College

More information

A Parallelization Design of JavaScript Execution Engine

A Parallelization Design of JavaScript Execution Engine , pp.171-184 http://dx.do.org/10.14257/mue.2014.9.7.15 A Parallelzaton Desgn of JavaScrpt Executon Engne Duan Huca 1,2, N Hong 2, Deng Feng 2 and Hu Lnln 2 1 Natonal Network New eda Engneerng Research

More information

[KV99] M. Kaul and R, Vemuri. Integrated Block-Processing and Design-Space Exploration in Temporal Partitioning for RTR Architectures.

[KV99] M. Kaul and R, Vemuri. Integrated Block-Processing and Design-Space Exploration in Temporal Partitioning for RTR Architectures. [KV99] M. Kaul and R, Vemur. Integrated Block-Processng and Desgn-Space Exploraton n Temporal Parttonng for RTR Archtectures. Reconfgurable Archtectures Workshop Proceedngs, Puerto Rco, Aprl 1999. [OCS98]

More information

Repeater Insertion for Two-Terminal Nets in Three-Dimensional Integrated Circuits

Repeater Insertion for Two-Terminal Nets in Three-Dimensional Integrated Circuits Repeater Inserton for Two-Termnal Nets n Three-Dmensonal Integrated Crcuts Hu Xu, Vasls F. Pavlds, and Govann De Mchel LSI - EPFL, CH-5, Swtzerland, {hu.xu,vasleos.pavlds,govann.demchel}@epfl.ch Abstract.

More information

Alternating Direction Method of Multipliers Implementation Using Apache Spark

Alternating Direction Method of Multipliers Implementation Using Apache Spark Alternatng Drecton Method of Multplers Implementaton Usng Apache Spark Deterch Lawson June 4, 2014 1 Introducton Many applcaton areas n optmzaton have benefted from recent trends towards massve datasets.

More information

SCALABLE AND VISUALIZATION-ORIENTED CLUSTERING FOR EXPLORATORY SPATIAL ANALYSIS

SCALABLE AND VISUALIZATION-ORIENTED CLUSTERING FOR EXPLORATORY SPATIAL ANALYSIS SCALABLE AND VISUALIZATION-ORIENTED CLUSTERING FOR EXPLORATORY SPATIAL ANALYSIS J.H.Guan, F.B.Zhu, F.L.Ban a School of Computer, Spatal Informaton & Dgtal Engneerng Center, Wuhan Unversty, Wuhan, 430079,

More information