Enabling GPU Virtualization in Cloud Environments

Size: px
Start display at page:

Download "Enabling GPU Virtualization in Cloud Environments"

Transcription

1 Enablng GPU Vrtualzaton n Cloud Envronments Sergo Iserte, Francsco J. Clemente-Castelló, Adrán Castelló, Rafael Mayo and Enrque S. Quntana-Ortí Department of Computer Scence and Engneerng, Unverstat Jaume I, Castelló de la Plana, Span Keywords: Abstract: Cloud Computng, GPU Vrtualzaton, Amazon Web Servces (AWS), OpenStack, Resource Management. The use of accelerators, such as graphcs processng unts (GPUs), to reduce the executon tme of computentensve applcatons has become popular durng the past few years. These devces ncrement the computatonal power of a node thanks to ther parallel archtecture. Ths trend has led cloud servce provders as Amazon or mddlewares such as OpenStack to add vrtual machnes (VMs) ncludng GPUs to ther facltes nstances. To fulfll these needs, the guest hosts must be equpped wth GPUs whch, unfortunately, wll be barely utlzed f a non GPU-enabled VM s runnng n the host. The soluton presented n ths work s based on GPU vrtualzaton and shareablty n order to reach an equlbrum between servce supply and the applcatons demand of accelerators. Concretely, we propose to decouple real GPUs from the nodes by usng the vrtualzaton technology rcuda. Wth ths software confguraton, GPUs can be accessed from any VM avodng the need of placng a physcal GPUs n each guest host. Moreover, we study the vablty of ths approach usng a publc cloud servce confguraton, and we develop a module for OpenStack n order to add support for the vrtualzed devces and the logc to manage them. The results demonstrate ths s a vable confguraton whch adds flexblty to current and well-known cloud solutons. 1 INTRODUCTION Nowadays, many cloud vendors have started offerng vrtual machnes (VMs) wth graphcs processng unts (GPUs) n order to provde GPGPU (general-purpose GPU) computaton servces. A few relevant examples nclude Amazon Web Servces (AWS) 1, Pengun Computng 2, Softlayer 3 and Mcrosoft Azure 4. In the publc scope, one of the most popular cloud vendors s AWS, whch offers a wde range of preconfgured nstances ready to be launched. Alternatvely, ownng the proper nfrastructure, a prvate cloud can be deployed usng a specfc mddleware such as OpenStack 5 or Opennebula 6. Unfortunately, sharng GPU resources among multple VMs n cloud envronments s more complex than n physcal servers. On one hand, nstances n publc clouds are not easly customzable. On the other, although the nstances n a prvate cloud can be customzed n many aspects, when referrng to GPUs the number of optons s reduced. As a result, nether vendors nor tools offer GPGPU servces. Remote vrtualzaton has been recently proposed to deal wth the low-usage problem. Some relevant examples nclude rcuda (Peña, 2013), DS- CUDA (Kawa et al., 2012), gvrtus (Gunta et al., 2010), vcuda (Sh et al., 2012), VOCL (Xao et al., 2012), and SnuCL (Km et al., 2012). Roughly speakng these vrtualzaton frameworks enable cluster confguratons wth fewer GPUs than nodes. The goal s that GPU-equpped nodes act as GPGPU servers, yeldng a GPU-sharng soluton that potentally acheves a hgher overall utlzaton of the accelerators n the system. The man goals of ths work are to study current cloud solutons n an HPC GPU-enabled scenaro, and to analyze and mprove them by addng flexblty va GPU vrtualzaton. In order to reach ths goal, we select rcuda, a vrtualzaton tool that s possbly the more complete and up-to-date for NVIDIA GPUs. The rest of the paper s structured as follows. In Secton 2 we ntroduce the technologes used n ths work; Secton 3 summarzes related work; the effort 249 Iserte, S., Clemente-Castelló, F., Castelló, A., Mayo, R. and Quntana-Ortí, E. Enablng GPU Vrtualzaton n Cloud Envronments. In Proceedngs of the 6th Internatonal Conference on Cloud Computng and Servces Scence (CLOSER 2016) - Volume 2, pages ISBN: Copyrght c 2016 by SCITEPRESS Scence and Technology Publcatons, Lda. All rghts reserved

2 CLOSER th Internatonal Conference on Cloud Computng and Servces Scence to use AWS s explaned n Secton 4; whle the work wth Openstack s descrbed n Secton 5; fnally, Secton 6 summarzes the advances and Secton 7 outlnes the next steps of ths research. 2 BACKGROUND 2.1 The rcuda Framework rcuda (Peña et al., 2014) s a mddleware that enables transparent access to any NVIDIA GPU devce present n a cluster from all compute nodes. The GPUs can also be shared among nodes, and a sngle node can use all the graphc accelerators as f they were local. rcuda s structured followng a clent-server dstrbuted archtecture and ts clent exposes the same nterface as the regular NVIDIA CUDA 6.5 release (NVIDIA Corp., ).Wth ths mddleware, applcatons are not aware that they are executed on top of a vrtualzaton layer. To deal wth new GPU programmng models, rcuda has been recently extended to accommodate drectvebased models such as OmpSs (Castelló et al., 2015a) and OpenACC (Castelló et al., 2015b). The ntegraton of remote GPGPU vrtualzaton wth global resource schedulers such as SLURM (Iserte et al., 2014) completes ths appealng technology, makng accelerator-enabled clusters more flexble and energyeffcent (Castelló et al., 2014). 2.2 Amazon Web Servces AWS (Amazon Web Servces, 2015) s a publc cloud computng provder, composed of several servces, such as cloud-based computaton, storage and other functonalty, that enables organzatons and/or ndvduals to deploy servces and applcatons on demand. These servces replace company-owned local IT nfrastructure and provde aglty and nstant elastcty matchng perfectly wth enterprse software requrements. From the pont of vew of HPC, AWS offers hgh performance facltes va nstances equpped wth GPUs and hgh performance network nterconnecton. 2.3 OpenStack OpenStack (OpenStack Foundaton, 2015) s a cloud operatng system (OS) that provdes Infrastructure as a Servce (IaaS). OpenStack controls large pools of compute, storage, and networkng resources throughout a datacenter. All these resources are managed through a dashboard or an API that gves admnstrators control whle empowerng ther users to provson resources through a web nterface or a command-lne nterface. OpenStack supports most recent hypervsors and handles provsonng and lfe-cycle management of VMs. The OpenStack archtecture offers flexblty to create a custom cloud, wth no propretary hardware or software requrements, and the ablty to ntegrate wth legacy systems and thrd party technologes. From the HPC perspectve, OpenStack offers hgh performance vrtual machne confguratons wth dfferent hardware archtectures. Even though n Open- Stack t s possble to work wth GPUs, the Nova project does not support ths archtecture yet. 3 STATE-OF-THE-ART Our solutons to the defcences exposed n the prevous secton reles on GPU vrtualzaton, sharng resources n order to attan a far balance between supply and demand. Whle several efforts wth the same goal have been ntated n the past, as exposed next, none of them s as ambtous as ours. The work n (Younge et al., 2013) allows the VM managed by the Xen hypervsor to access the GPUs n a physcal node, but wth ths soluton a node cannot use more GPUs than those locally hosted, and an dle GPU cannot be shared wth other nodes. The soluton presented by gvrtus (Gunta et al., 2010) vrtualzes GPUs and makes them accessble for any VM n the cluster. However, ths knd of vrtualzaton strongly depends on the hypervsor, and so does ts performance. Smlar soluton s presented n gcloud (Dab et al., 2013). Whle ths soluton s not yet ntegrated n a Cloud Computng Manager, ts man drawback s that the applcaton s code must be modfed n order to run n the vrtual-gpu envronment. A runtme component to provde abstracton and sharng of GPUs s presented n (Becch et al., 2012), whch allows schedulng polces to solate and share GPUs n a cluster for a set of applcatons. The work ntroduced n (Jun et al., 2014) s more mature; however, t s only focused on compute-ntensve HPC applcatons. Our proposal goes further, not only brngng solutons for all knd of HPC applcatons, but also amng to boost flexblty n the use of GPUs. 250

3 Enablng GPU Vrtualzaton n Cloud Envronments 4 GPGPU SHARE SERVICE WITH AWS 4.1 Current Features Instances An nstance s a pre-confgured VM focused on an specfc target. Among the large lst of nstances offered by AWS, we can fnd specalzed versons for general-purpose (T2, M4 and M3); computer scence (C4 and C3); memory (R3); storage (I2 and D2) and GPU capable (G2). Each type of nstance has ts own purpose and cost (prce). Moreover, each type offers a dfferent number of CPUs as well as network nterconnecton, whch can be: low, medum, hgh or 10Gb. For our study, we worked n the AWS avalablty zone US EAST (N. VIRGINIA). The nstances avalable n that case present the features reported n Table 1. Table 1: Shown HPC nstances avalable n US EAST (N. VIRGINIA) n June Name vcpus Memory Network GPUs Prce c3.2xlarge 8 15 GB Hgh 0 $ 42 c3.8xlarge GB 10 Gb 0 $ 1.68 g2.2xlarge 8 15 GB Hgh 1 $ 65 g2.8xlarge GB 10 Gb 4 $ 2.6 For the followng experments, we select C3 famly nstances, whch are not equpped wth GPUs, as clents; whereas nstances of the G2 famly wll act as GPU-enabled servers Networkng Table 1 shows that each nstance ntegrates a dfferent network. As the bandwdth s crtcal when GPU vrtualzaton s appled, we frst perform a smple test to verfy the real network bandwdth. Table 2: IPERF results between selected nstances. Server Clent Network Bandwdth g2.8xlarge c3.2xlarge Hgh 1 Gb/s g2.8xlarge c3.8xlarge 10Gb 7.5 Gb/s g2.8xlarge g2.2xlarge Hgh 1 Gb/s g2.8xlarge g2.8xlarge 10Gb 7.5 Gb/s To evaluate the actual bandwdth, we executed the IPERF 7 tool between the nstances descrbed n Ta- 7 ble 1, wth the results shown n Table 2. From ths experment, we can derve that network Hgh corresponds to a 1 Gb nterconnect whle 10 Gb has a real bandwdth of 7.5 Gb/s. Moreover, t seems that the bandwdth of the nstances equpped wth a Hgh nterconnecton network s constraned by software to 1 Gb/s snce the theoretcal and real bandwdth match perfectly. The real gap between sustaned and theoretcal bandwdth can be observed wth the 10 Gb nterconnecton, whch reaches up to 7.5 Gb/s GPUs An nstance reles on a VM that runs on a real node wth ts own vrtualzed components. Therefore AWS can leverage a vrtualzaton framework to offer GPU servces to all the nstances. Although the nvda-sm command ndcates that the GPUs nstalled are NVIDIA GRID K520, we need to verfy that these are non-vrtualzed devces. For ths purpose, we execute the NVIDIA SDK bandwdthtest. As shown n Table 3, the bandwdth acheved n ths test s hgher than the network bandwdth, whch suggests that the accelerator s an actual GPU. Table 3: Results of bandwdthtest transferrng 32MB usng pageable memory n a local GPU. Name Data Movement Bandwdth g2.2xlarge Host to Devce 3,004 MB/s g2.2xlarge Devce to Host 2,809 MB/s g2.8xlarge Host to Devce 2,814 MB/s g2.8xlarge Devce to Host 3,182 MB/s 4.2 Testbed Scenaros All scenaros are based on the maxmum number of nstances that a user can freely select wthout submttng a formal request. In partcular, the maxmum number for g2.2xlarge s 5; for g2.8xlarge t s 2. Ant the nstances operate the RHEL bt OS and verson 6.5 of CUDA. We desgn three confguraton scenaros for our tests: Scenaro A (Fgure 1(a)) shows a common confguraton n GPU-accelerated clusters, wth each node populated wth a sngle GPU. Here, a node can access 5 GPUs usng the Hgh network. Scenaro B (Fgure 1(b)) s composed of 2 server nodes, equpped wth 4 GPUs each, and a GPUless clent. Ths scenaro ncludes a 10Gb network, and the clent can execute the applcaton usng up to 8 GPUs. 251

4 CLOSER th Internatonal Conference on Cloud Computng and Servces Scence Scenaro C (Fgure 1(c)) combnes scenaros A and B. A sngle clent, usng a 1Gb network nterconnecton, can leverage 13 GPUs as f they were local. Once the scenaros are confgured from the pont of vew of hardware, the rcuda mddleware needs to be nstalled n order to add the requred flexblty to the system. The rcuda server s executed n the server nodes and the rcuda lbrares are nvoked from the node that acts as clent. In order to evaluate the network bandwdth usng a remote GPU, we re-appled NVIDIA SDK bandwdthtest. Table 4 exposes that the bandwdth s lmted by the network. Table 4: Results of bandwdthtest transferrng 32MB usng pageable memory n a remote GPU usng rcuda. Scenaro Data Movement Network Bandwdth A Host-to-Devce Hgh 127 MB/s A Devce-to-Host Hgh 126 MB/s B Host-to-Devce 10 Gb 858 MB/s B Devce-to-Host 10 Gb 843 MB/s 4.3 Expermental Results The frst applcaton s MonteCarloMultGPU, from the NVIDIA SDK, a code that s compute bound (ts executon barely nvolves memory operatons). Ths was launched wth the default confguraton, scalng=weak, whch adjusts the sze of the problem dependng on the number of accelerators. Fgure 2 depcts the optons per second calculated by the applcaton runnng on the scenaros n Fgure 1 as well as usng local GPUs. For clarty, we have removed the results observed for Scenaro B as they are exactly the same as those obtaned from Scenaro C wth up to 8 GPUs. In ths partcular case, rcuda (remote GPUs) outperforms CUDA (local GPUs) because the former loads the lbrares when the daemon s started (Peña, 2013). Wth rcuda we can observe dfferences n the results between both scenaros. Here, Scenaro A can ncrease the throughput because the GPUs do not share the PCI bus wth other devces as each node only s equpped wth one GPU. On the other hand, when the 4-GPU nstances ( g2.8xlarge ) are added (Scenaro C), the PCI bus constrans the communcaton bandwdth, hurtng the scalablty. The second applcaton, LAMMPS 8, s a classcal molecular dynamcs smulator that can be appled at the atomc, meso, or contnuum scale. From the mplementaton perspectve, ths mult-process applcaton employs at least one GPU to host ts processes, but can beneft from the presence of multple GPUs. Fgure 3(a) shows that, for ths applcaton, the use of remote GPUs does not offer any advantage over the orgnal CUDA. Furthermore, for the executon on remote GPUs, the dfference between both networks s small, although, the results observed wth the Hgh network are worse than those obtaned wth the 10 Gb network. In executon of LAMMPS on a larger problem (see Fgure 3(b)), CUDA stll performs better, but the nterestng pont s the executon tme when usng remote GPUs. These are almost the same even wth dfferent networks, whch ndcates that the transfers turn the nterconnecton network nto a bottleneck. For ths type of applcaton, enlargng the problem sze compensates the negatve effect of a slower network. 4.4 Dscusson The prevous experments reveal that the AWS GPUnstances are not approprate for HPC because nether the network nor the accelerators are powerful enough to delver hgh performance when runnng compute-ntensve parallel applcatons. As (Peña, 2013) demonstrates, network and devce types are crtcal factors to performance. In other words, AWS s more orented toward offerng a general-purpose servce than to provde hgh performance. Also, AWS fals n offerng flexblty as t enforces the user to choose between a basc nstance wth a GPU and a powerful nstance wth 4 GPUs. Table 1 shows that the resources of the g2.8xlarge are quadrupled, but so s the cost per hour. Therefore, n the case of havng other necesstes (nstances type), usng GPU vrtualzaton technology we could n prncple attach an accelerator to any type of nstance. Furthermore, reducng the budget spent n cloud servces s possble by customzng the resources of the avalable nstances. For example, we can work on an nstance wth 2 GPUs for only $ 1.3 by launchng 2 g2.2xlarge and usng remote GPUs, avodng to pay the double for features that we do not need n g2.8xlarge. In terms of GPU-shareablty, AWS reserves GPU-capable physcal machnes whch wll be watng for a GPU-nstance request. Investng n expensve accelerators to keep them n a standby state s counter-productve. It makes more sense to dedcate less devces, accessble from any machne, resultng n a hgher utlzaton rate

5 Enablng GPU Vrtualzaton n Cloud Envronments 4GPUs 4GPUs 1Gb 10Gb 1Gb 4GPUs 4GPUs (a) 5 remote GPUs over 1GbE. (b) 8 remote GPUs over 10GbE. (c) 13 remote GPUs over 1GbE. Fgure 1: Confguratons where an AWS nstance, actng as a clent, employs several remote GPUs. 5 GPGPU SHARED SERVICE IN OPENSTACK 5.1 Managng Remote GPUs wth OpenStack Fgure 2: NVIDIA SDK MonteCarloMultGPU executon usng local (CUDA) and remote (Scenaros A and C) GPUs. (a) Run sze = 10 The man dea s to evolve from the orgnal OpenStack archtecture (see Fgure 4(a)) to a soluton where a new shared servce, responsble for the GPU accelerators, becomes ntegrated nto the archtecture (see Fgure 4(b)). Ths new servce brngs more flexblty when managng GPUs and new workng modes for GPGPU computaton n the Cloud. As llustrated n Fgure 5, we alter the orgnal OpenStack Dashboard wth a new parser, whch splts the HTTP query n order to make use of both the GPGPU API for GPU-related operatons, and the Nova API for the rest of the computatons. The new GPGPU Servce grants access to GPUs n a VM, but also allows the creaton of GPU-pools, consstng of a set of ndependent GPUs (logcally dsattached from the nodes) that can be assgned to one or more VMs. Thanks to the modular approach of the GPGPU Servce, the Nova Project does not need to be modfed, and the tool can be easly ported to other Cloud Computng solutons. 5.2 Workng Modes (b) Run sze = 200 Fgure 3: Executon tme of LAMMPS. The developed module allows users to set up any of the scenaros dsplayed n Fgure 6. The users are gven two confguraton optons to decde whether a physcal GPU wll be completely reserved for an nstance (frst column) or the nstance wll address a partton of the GPU as f t were a real devce (second column). We refer to ths as the mode, wth possble values beng: exclusve or shared. Let us assume there are 4 GPU devces n the cluster (ndependently of where they were hosted). An example of these scenaros s shown n Fgure 6. There, n the exclusve 253

6 CLOSER th Internatonal Conference on Cloud Computng and Servces Scence (a) Verson Icehouse Fgure 6: Examples of workng modes. (b) Wth GPGPU module Fgure 4: OpenStack Archtecture. Fgure 5: Internal Communcaton among modules. GPUPool mode the nstance monopolzes all the GPUs; whle n the shared mode, the GPUs are parttoned. As a result of sharng the GPU memory, the nstance wll be able to work wth up to 8 GPUs, provded that each partton can be addressed as an ndependent GPU. Moreover, the users are also responsble for decdng whether a GPU (or a pool) wll be assgned to other nstances. Ths behavor s refereed as scope, and t determnes that a group of nstances s logcally connected to a pool of GPUs. Workng wth the publc scope (bottom row of Fgure 6) mples that the GPUs of a pool can be used smultaneously by all the nstances lnked to t. Agan, the GPU pool can be composed of exclusve or shared GPUs. 5.3 GPUsMemor y( MB) 2048 Publ cpool Ex c l us vegpus Fgure 7: Launchng Instances and assgnng GPUs. nstance wthout accelerators of ths knd. When the opton New GPU Pool s chosen, felds for the pool confguraton would appear (see Fgure 7). Furthermore, a new panel wth the exstent GPUs dsplays all the nformaton related to GPUs (see Fgure 8). User Interface In order to deal wth the new features, several modfcatons have been planned n the OpenStack Dashboard, they have not been mplemented yet, though. Frst of all, the Instance Launch Panel should be extended wth a new feld, where the user could assgn an exstng GPU pool, create a new one, or keep the 254 New GPUCount GPUs I D Addr es s Memor y Sc ope PoolorI ns t anc e Mode 1 1 1: Publ c Pool1 Shar ed 2 1 1: Publ c Pool2 Shar ed 3 1 2: Pr vat e ns t anc e00362 Pr vat e I ns t anc es GPUs Fgure 8: GPU Informaton Panel. Act ons

7 Enablng GPU Vrtualzaton n Cloud Envronments 5.4 Expermental Results Followng tests were executed on 2 sets of nodes usng a 1Gb Ethernet network. All the nodes composed by an Intel Xeon E7420 quadcore processor, at 2.13 GHz, and 16 GB DDR2 RAM at 667 MHz. The frst set, n charge of provdng the cloud envronment, conssted of 3 nodes. To deploy an IaaS, we used OpenStack Icehouse verson, and QEMU/KVM 12.1 as the hypervsor. A fully-featured Open- Stack deployment requres at least three nodes: a controller manages the compute nodes where the VMs are hosted; a network node manages the logc vrtual network for the VMs, and one or more compute nodes run the hypervsor and VMs. The second set, composed of 4 nodes, were auxlary servers wth a Tesla C1060 GPU each. The OS was a Centos 6.6; the GPUs used CUDA 6.5; and rcuda v5.0 as GPU vrtualzaton framework. We have desgned 6 dfferent set-ups whch can be dvded nto 2 groups: exclusve and shared GPUs. The exclusve mode provdes, at most, 4 accelerators. The number of avalable GPU n shared mode wll depend on the partton sze. In ths case, we halved the GPU memory, resultng 8 parttons that can be addressed as ndependent GPUs. For each group, we deployed vrtual clusters of 1, 2 and 3 nodes, where the applcaton processes were executed. The nstances were launched wth the OpenStack predefned flavor m1.medum, whch determnes a confguraton of VMs consstng of 2 cores and 4 GB of memory. MCUDA-MEME (Lu et al., 2010) was the applcaton selected to test the set-ups. Thus s an MPI software, where each process must have access to a GPU. Therefore, the number of GPUs determnes the maxmum number of processes we can launch. Fgure 9 compares the executon tme of the applcaton wth dfferent confguratons over dfferent setups. We used up to 3 nodes to spread the processes and launched up to 8 processes (only 4 n exclusve mode), one per remote GPU. We can frst observe that the performance s hgher wth more than one node, because the traffc network s dstrbuted among the nodes. In addton, the shortest executon tme s obtaned by both modes (exclusve and shared) when runnng ther maxmum number of processes wth more than one node. Ths seems to mply that t s not worth to scale (ncrease) the number of resources, because the performance growth rate s barely ncreasng. Although, the tme s lower when the GPUs are shared, the setup cannot take advantage of an ncrease n the number of devces. Fgure 9: Scalablty results of MCUDAMEME wth a dfferent number of MPI processes. 5.5 Dscusson The network nterconnect restrcts the performance of our executons. The analyss n (Peña, 2013) reveals that mprovng the network nfrastructure can make a bg dfferent for GPU vrtualzaton. The most remarkable achevement s the wde range of possble confguratons and the flexblty to adapt a system to ft the user requrements. In addton, wth ths vrtualzaton technology, the requests for GPU devces can be fulflled wth small nvestment n nfrastructure and mantenance. Energy can be saved not only thanks to the remote access and the ablty to emulate several GPUs usng only a few real ones, but also by consoldatng the accelerators n a sngle machne (when possble), or turnng down nodes when ther GPUs are dle. 6 CONCLUSIONS We have presented a complete study of the possbltes offered by AWS when t comes to GPUs. The constrants mposed by ths servce motvated us to deploy our own prvate cloud, n order to gan flexblty when dealng wth these accelerators. For ths purpose, we have ntroduced an extenson of OpenStack whch can be easly exploted to create GPU-nstances as well as manage the physcal GPUs to better proft. As we expected, due to the lmted bandwdth of the nterconnects used n the expermentaton, the performances observed for the GPU vrtualzed scenaros n the tests were qute low. On the other hand, we have created new operaton modes that open nterestng new ways to leverage GPUs n stuatons where havng access to a GPU s more mportant than havng a powerful GPU to boost the performance. 255

8 CLOSER th Internatonal Conference on Cloud Computng and Servces Scence 7 FUTURE WORK The frst tem n the lst of pendng work s an upgrade of the network to an nterconnect that s more prone to HPC. In partcular, portng the setup and the tests to an nfrastructure wth an Infnband network wll shed lght on the vablty of ths knd of solutons. Smlar reasons, motvate us to try other Cloud vendors whch better support for HPC. Lookng for stuatons where performance s less mportant than flexblty wll drve us to explore alternatve tools to easly deploy GPU-programmng computer labs. Fnally, an nterestng future work s to desgn new strateges n order to decde where a remote GPUs s created and assgned to a physcal devce Concretely, to nnovate schedulng polces can enhance the flexblty offered by the GPGPU module for OpenStack. ACKNOWLEDGEMENTS The authors would lke to thank the IT members of the department Gustavo Edo and Vcente Roca for ther help. Ths research was supported by Unverstat Jaume I research project (P11B ); and project TIN R from MINECO and FEDER. The ntal verson of rcuda was jontly developed by Unverstat Poltècnca de Valènca (UPV) and Unverstat Jaume I de Castellón untl year 201 Ths ntal development was later splt nto two branches. Part of the UPV verson was used n ths paper and t was supported by Generaltat Valencana under Grants PROMETEO 2008/060 and Prometeo II 2013/009. REFERENCES Amazon Web Servces (2015). Amazon web servces. Accessed: Becch, M., Sajjapongse, K., Graves, I., Procter, A., Rav, V., and Chakradhar, S. (2012). A vrtual memory based runtme to support mult-tenancy n clusters wth GPUs. In 21st Int. symp. on Hgh-Performance Parallel and Dstrbuted Computng. Castelló, A., Duato, J., Mayo, R., Peña, A. J., Quntana- Ortí, E. S., Roca, V., and Slla, F. (2014). On the use of remote GPUs and low-power processors for the acceleraton of scentfc applcatons. In The Fourth Int. Conf. on Smart Grds, Green Communcatons and IT Energy-aware Technologes, pages 57 62, France. Castelló, A., Mayo, R., Planas, J., and Quntana-Ortí, E. S. (2015a). Explotng task-parallelsm on GPU clusters va OmpSs and rcuda vrtualzaton. In I IEEE Int. Workshop on Reengneerng for Parallelsm n Heterogeneous Parallel Platforms, Helsnk (Fnland). Castelló, A., Peña, A. J., Mayo, R., Balaj, P., and Quntana- Ortí, E. S. (2015b). Explorng the sutablty of remote GPGPU vrtualzaton for the OpenACC programmng model usng rcuda. In IEEE Int. Conference on Cluster Computng, Chcago, IL (USA). Dab, K. M., Rafque, M. M., and Hefeeda, M. (2013). Dynamc sharng of GPUs n cloud systems. In Parallel and Dstrbuted Processng Symp. Workshops & PhD Forum, 2013 IEEE 27th Internatonal. Gunta, G., Montella, R., Agrllo, G., and Covello, G. (2010). A GPGPU transparent vrtualzaton component for hgh performance computng clouds. In Euro- Par, Parallel Processng, pages Sprnger. Iserte, S., Castelló, A., Mayo, R., Quntana-Ortí, E. S., Reaño, C., Prades, J., Slla, F., and Duato, J. (2014). SLURM support for remote GPU vrtualzaton: Implementaton and performance study. In Int. Symposum on Computer Archtecture and Hgh Performance Computng, Pars, France. Jun, T. J., Van Quoc Dung, M. H. Y., Km, D., Cho, H., and Hahm, J. (2014). GPGPU enabled HPC cloud platform based on OpenStack. Kawa, A., Yasuoka, K., Yoshkawa, K., and Narum, T. (2012). Dstrbuted-shared CUDA: Vrtualzaton of large-scale GPU systems for programmablty and relablty. In The Fourth Int. Conf. on Future Computatonal Technologes and Applcatons, pages Km, J., Seo, S., Lee, J., Nah, J., Jo, G., and Lee, J. (2012). SnuCL: an OpenCL framework for heterogeneous CPU/GPU clusters. In Int. Conf. on Supercomputng (ICS). Lu, Y., Schmdt, B., Lu, W., and Maskell, D. L. (2010). CUDA-MEME: Acceleratng motf dscovery n bologcal sequences usng CUDA-enabled graphcs processng unts. Pattern Recognton Letters, 31(14). NVIDIA Corp. CUDA API Reference Manual Verson 6.5. OpenStack Foundaton (2015). OpenStack. Accessed: Peña, A. J. (2013). Vrtualzaton of accelerators n hgh performance clusters. PhD thess, Unverstat Jaume I, Castellon, Span. Peña, A. J., Reaño, C., Slla, F., Mayo, R., Quntana-Ortí, E. S., and Duato, J. (2014). A complete and effcent CUDA-sharng soluton for HPC clusters. Parallel Computer, 40(10). Sh, L., Chen, H., Sun, J., and L, K. (2012). vcuda: GPU-accelerated hgh-performance computng n vrtual machnes. IEEE Trans. on Comput., 61(6). Xao, S., Balaj, P., Zhu, Q., Thakur, R., Coghlan, S., Ln, H., Wen, G., Hong, J., and Feng, W. (2012). VOCL: An optmzed envronment for transparent vrtualzaton of graphcs processng unts. In Innovatve Parallel Computng. IEEE. Younge, A. J., Walters, J. P., Crago, S., and Fox, G. C. (2013). Enablng hgh performance computng n cloud nfrastructure usng vrtualzed GPUs. 256

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

Solution Brief: Creating a Secure Base in a Virtual World

Solution Brief: Creating a Secure Base in a Virtual World Soluton Bref: Creatng a Secure Base n a Vrtual World Soluton Bref: Creatng a Secure Base n a Vrtual World Abstract The adopton rate of Vrtual Machnes has exploded at most organzatons, drven by the mproved

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

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

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

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

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

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

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

an assocated logc allows the proof of safety and lveness propertes. The Unty model nvolves on the one hand a programmng language and, on the other han

an assocated logc allows the proof of safety and lveness propertes. The Unty model nvolves on the one hand a programmng language and, on the other han UNITY as a Tool for Desgn and Valdaton of a Data Replcaton System Phlppe Quennec Gerard Padou CENA IRIT-ENSEEIHT y Nnth Internatonal Conference on Systems Engneerng Unversty of Nevada, Las Vegas { 14-16

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

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

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

Motivation. EE 457 Unit 4. Throughput vs. Latency. Performance Depends on View Point?! Computer System Performance. An individual user wants to:

Motivation. EE 457 Unit 4. Throughput vs. Latency. Performance Depends on View Point?! Computer System Performance. An individual user wants to: 4.1 4.2 Motvaton EE 457 Unt 4 Computer System Performance An ndvdual user wants to: Mnmze sngle program executon tme A datacenter owner wants to: Maxmze number of Mnmze ( ) http://e-tellgentnternetmarketng.com/webste/frustrated-computer-user-2/

More information

Mellanox CloudX, Mirantis Fuel Solution Guide

Mellanox CloudX, Mirantis Fuel Solution Guide Mellanox CloudX, Mrants Fuel Soluton Gude Rev.0 www.mellanox.com NOTE: THIS HARDWARE, SOFTWARE OR TEST SUITE PRODUCT ( PRODUCT(S) ) AND ITS RELATED DOCUMENTATION ARE PROVIDED BY MELLANOX TECHNOLOGIES AS-IS

More information

ELEC 377 Operating Systems. Week 6 Class 3

ELEC 377 Operating Systems. Week 6 Class 3 ELEC 377 Operatng Systems Week 6 Class 3 Last Class Memory Management Memory Pagng Pagng Structure ELEC 377 Operatng Systems Today Pagng Szes Vrtual Memory Concept Demand Pagng ELEC 377 Operatng Systems

More information

The Greedy Method. Outline and Reading. Change Money Problem. Greedy Algorithms. Applications of the Greedy Strategy. The Greedy Method Technique

The Greedy Method. Outline and Reading. Change Money Problem. Greedy Algorithms. Applications of the Greedy Strategy. The Greedy Method Technique //00 :0 AM Outlne and Readng The Greedy Method The Greedy Method Technque (secton.) Fractonal Knapsack Problem (secton..) Task Schedulng (secton..) Mnmum Spannng Trees (secton.) Change Money Problem Greedy

More information

Assembler. Building a Modern Computer From First Principles.

Assembler. Building a Modern Computer From First Principles. Assembler Buldng a Modern Computer From Frst Prncples www.nand2tetrs.org Elements of Computng Systems, Nsan & Schocken, MIT Press, www.nand2tetrs.org, Chapter 6: Assembler slde Where we are at: Human Thought

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

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

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

Internet Traffic Managers

Internet Traffic Managers Internet Traffc Managers Ibrahm Matta matta@cs.bu.edu www.cs.bu.edu/faculty/matta Computer Scence Department Boston Unversty Boston, MA 225 Jont work wth members of the WING group: Azer Bestavros, John

More information

Mathematics 256 a course in differential equations for engineering students

Mathematics 256 a course in differential equations for engineering students Mathematcs 56 a course n dfferental equatons for engneerng students Chapter 5. More effcent methods of numercal soluton Euler s method s qute neffcent. Because the error s essentally proportonal to the

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

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

For instance, ; the five basic number-sets are increasingly more n A B & B A A = B (1)

For instance, ; the five basic number-sets are increasingly more n A B & B A A = B (1) Secton 1.2 Subsets and the Boolean operatons on sets If every element of the set A s an element of the set B, we say that A s a subset of B, or that A s contaned n B, or that B contans A, and we wrte A

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

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

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

A Resources Virtualization Approach Supporting Uniform Access to Heterogeneous Grid Resources 1

A Resources Virtualization Approach Supporting Uniform Access to Heterogeneous Grid Resources 1 A Resources Vrtualzaton Approach Supportng Unform Access to Heterogeneous Grd Resources 1 Cunhao Fang 1, Yaoxue Zhang 2, Song Cao 3 1 Tsnghua Natonal Labatory of Inforamaton Scence and Technology 2 Department

More information

Improved Energy-Efficiency in Cloud Datacenters with Interference-Aware Virtual Machine Placement

Improved Energy-Efficiency in Cloud Datacenters with Interference-Aware Virtual Machine Placement Improved Energy-Effcency n Cloud Datacenters wth Interference-Aware Vrtual Machne Placement Ismael Sols Moreno 1, Renyu Yang 2, Je Xu 1, 2, Tanyu Wo 2 School of Computng 1 Unversty of Leeds Leeds, UK {scsm,

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

Evaluation of an Enhanced Scheme for High-level Nested Network Mobility

Evaluation of an Enhanced Scheme for High-level Nested Network Mobility IJCSNS Internatonal Journal of Computer Scence and Network Securty, VOL.15 No.10, October 2015 1 Evaluaton of an Enhanced Scheme for Hgh-level Nested Network Moblty Mohammed Babker Al Mohammed, Asha Hassan.

More information

IP Camera Configuration Software Instruction Manual

IP Camera Configuration Software Instruction Manual IP Camera 9483 - Confguraton Software Instructon Manual VBD 612-4 (10.14) Dear Customer, Wth your purchase of ths IP Camera, you have chosen a qualty product manufactured by RADEMACHER. Thank you for the

More information

VRT012 User s guide V0.1. Address: Žirmūnų g. 27, Vilnius LT-09105, Phone: (370-5) , Fax: (370-5) ,

VRT012 User s guide V0.1. Address: Žirmūnų g. 27, Vilnius LT-09105, Phone: (370-5) , Fax: (370-5) , VRT012 User s gude V0.1 Thank you for purchasng our product. We hope ths user-frendly devce wll be helpful n realsng your deas and brngng comfort to your lfe. Please take few mnutes to read ths manual

More information

Lobachevsky State University of Nizhni Novgorod. Polyhedron. Quick Start Guide

Lobachevsky State University of Nizhni Novgorod. Polyhedron. Quick Start Guide Lobachevsky State Unversty of Nzhn Novgorod Polyhedron Quck Start Gude Nzhn Novgorod 2016 Contents Specfcaton of Polyhedron software... 3 Theoretcal background... 4 1. Interface of Polyhedron... 6 1.1.

More information

Game Based Virtual Bandwidth Allocation for Virtual Networks in Data Centers

Game Based Virtual Bandwidth Allocation for Virtual Networks in Data Centers Avaable onlne at www.scencedrect.com Proceda Engneerng 23 (20) 780 785 Power Electroncs and Engneerng Applcaton, 20 Game Based Vrtual Bandwdth Allocaton for Vrtual Networks n Data Centers Cu-rong Wang,

More information

Virtual Machine Placement Based on the VM Performance Models in Cloud

Virtual Machine Placement Based on the VM Performance Models in Cloud Vrtual Machne Placement Based on the VM Performance Models n Cloud Hu Zhao, Qnghua Zheng, Member, IEEE, Wezhan Zhang Member, IEEE, Yuxuan Chen, Yunhu Huang SPKLSTN Lab, Department of Computer Scence and

More information

Wavefront Reconstructor

Wavefront Reconstructor A Dstrbuted Smplex B-Splne Based Wavefront Reconstructor Coen de Vsser and Mchel Verhaegen 14-12-201212 2012 Delft Unversty of Technology Contents Introducton Wavefront reconstructon usng Smplex B-Splnes

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

Related-Mode Attacks on CTR Encryption Mode

Related-Mode Attacks on CTR Encryption Mode Internatonal Journal of Network Securty, Vol.4, No.3, PP.282 287, May 2007 282 Related-Mode Attacks on CTR Encrypton Mode Dayn Wang, Dongda Ln, and Wenlng Wu (Correspondng author: Dayn Wang) Key Laboratory

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 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 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

Complex Numbers. Now we also saw that if a and b were both positive then ab = a b. For a second let s forget that restriction and do the following.

Complex Numbers. Now we also saw that if a and b were both positive then ab = a b. For a second let s forget that restriction and do the following. Complex Numbers The last topc n ths secton s not really related to most of what we ve done n ths chapter, although t s somewhat related to the radcals secton as we wll see. We also won t need the materal

More information

An Approach to Optimized Resource Scheduling Algorithm for Open-source Cloud Systems

An Approach to Optimized Resource Scheduling Algorithm for Open-source Cloud Systems The Ffth Annual ChnaGrd Conference An Approach to Optmzed Resource Schedulng Algorthm for Open-source Cloud Systems Ha Zhong 1, 2, Kun Tao 1, Xueje Zhang 1, 2 1 School of Informaton Scence and Engneerng,

More information

Circuit Analysis I (ENGR 2405) Chapter 3 Method of Analysis Nodal(KCL) and Mesh(KVL)

Circuit Analysis I (ENGR 2405) Chapter 3 Method of Analysis Nodal(KCL) and Mesh(KVL) Crcut Analyss I (ENG 405) Chapter Method of Analyss Nodal(KCL) and Mesh(KVL) Nodal Analyss If nstead of focusng on the oltages of the crcut elements, one looks at the oltages at the nodes of the crcut,

More information

CMPS 10 Introduction to Computer Science Lecture Notes

CMPS 10 Introduction to Computer Science Lecture Notes CPS 0 Introducton to Computer Scence Lecture Notes Chapter : Algorthm Desgn How should we present algorthms? Natural languages lke Englsh, Spansh, or French whch are rch n nterpretaton and meanng are not

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

Parallelization of a Series of Extreme Learning Machine Algorithms Based on Spark

Parallelization of a Series of Extreme Learning Machine Algorithms Based on Spark Parallelzaton of a Seres of Extreme Machne Algorthms Based on Spark Tantan Lu, Zhy Fang, Chen Zhao, Yngmn Zhou College of Computer Scence and Technology Jln Unversty, JLU Changchun, Chna e-mal: lutt1992x@sna.com

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

Research Article Adaptive Cost-Based Task Scheduling in Cloud Environment

Research Article Adaptive Cost-Based Task Scheduling in Cloud Environment Scentfc Programmng Volume 2016, Artcle ID 8239239, 9 pages http://dx.do.org/10.1155/2016/8239239 Research Artcle Adaptve Cost-Based Task Schedulng n Cloud Envronment Mohammed A. S. Mosleh, 1 G. Radhaman,

More information

Analysis of Collaborative Distributed Admission Control in x Networks

Analysis of Collaborative Distributed Admission Control in x Networks 1 Analyss of Collaboratve Dstrbuted Admsson Control n 82.11x Networks Thnh Nguyen, Member, IEEE, Ken Nguyen, Member, IEEE, Lnha He, Member, IEEE, Abstract Wth the recent surge of wreless home networks,

More information

A QoS-aware Scheduling Scheme for Software-Defined Storage Oriented iscsi Target

A QoS-aware Scheduling Scheme for Software-Defined Storage Oriented iscsi Target A QoS-aware Schedulng Scheme for Software-Defned Storage Orented SCSI Target Xanghu Meng 1,2, Xuewen Zeng 1, Xao Chen 1, Xaozhou Ye 1,* 1 Natonal Network New Meda Engneerng Research Center, Insttute of

More information

Application of Improved Fish Swarm Algorithm in Cloud Computing Resource Scheduling

Application of Improved Fish Swarm Algorithm in Cloud Computing Resource Scheduling , pp.40-45 http://dx.do.org/10.14257/astl.2017.143.08 Applcaton of Improved Fsh Swarm Algorthm n Cloud Computng Resource Schedulng Yu Lu, Fangtao Lu School of Informaton Engneerng, Chongqng Vocatonal Insttute

More information

Analysis of Continuous Beams in General

Analysis of Continuous Beams in General Analyss of Contnuous Beams n General Contnuous beams consdered here are prsmatc, rgdly connected to each beam segment and supported at varous ponts along the beam. onts are selected at ponts of support,

More information

A dynamic bandwidth allocator for virtual machines in a cloud environment

A dynamic bandwidth allocator for virtual machines in a cloud environment A dynamc bandwdth allocator for vrtual machnes n a cloud envronment Ahmed Amamou, Manel Bourguba, Kamel Haddadou, Guy Pujolle To cte ths verson: Ahmed Amamou, Manel Bourguba, Kamel Haddadou, Guy Pujolle.

More information

Scalability of a Mobile Cloud Management System

Scalability of a Mobile Cloud Management System Scalablty of a Moble Cloud Management System Roberto Bfulco Unversty of Napol Federco II roberto.bfulco2@unna.t Marcus Brunner NEC Laboratores Europe brunner@neclab.eu Peer Hasselmeyer NEC Laboratores

More information

A KIND OF ROUTING MODEL IN PEER-TO-PEER NETWORK BASED ON SUCCESSFUL ACCESSING RATE

A KIND OF ROUTING MODEL IN PEER-TO-PEER NETWORK BASED ON SUCCESSFUL ACCESSING RATE A KIND OF ROUTING MODEL IN PEER-TO-PEER NETWORK BASED ON SUCCESSFUL ACCESSING RATE 1 TAO LIU, 2 JI-JUN XU 1 College of Informaton Scence and Technology, Zhengzhou Normal Unversty, Chna 2 School of Mathematcs

More information

TripS: Automated Multi-tiered Data Placement in a Geo-distributed Cloud Environment

TripS: Automated Multi-tiered Data Placement in a Geo-distributed Cloud Environment TrpS: Automated Mult-tered Data Placement n a Geo-dstrbuted Cloud Envronment Kwangsung Oh, Abhshek Chandra, and Jon Wessman Department of Computer Scence and Engneerng Unversty of Mnnesota Twn Ctes Mnneapols,

More information

4/11/17. Agenda. Princeton University Computer Science 217: Introduction to Programming Systems. Goals of this Lecture. Storage Management.

4/11/17. Agenda. Princeton University Computer Science 217: Introduction to Programming Systems. Goals of this Lecture. Storage Management. //7 Prnceton Unversty Computer Scence 7: Introducton to Programmng Systems Goals of ths Lecture Storage Management Help you learn about: Localty and cachng Typcal storage herarchy Vrtual memory How the

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

Agile Data Streaming for Grid Applications

Agile Data Streaming for Grid Applications Agle Data Streamng for Grd Applcatons Wen Zhang, Junwe Cao 2,3*, Ysheng Zhong,3, Lanchen Lu,3, and Cheng Wu,3 Department of Automaton, Tsnghua Unversty, Bejng 00084, Chna 2 Research Insttute of Informaton

More information

3D vector computer graphics

3D vector computer graphics 3D vector computer graphcs Paolo Varagnolo: freelance engneer Padova Aprl 2016 Prvate Practce ----------------------------------- 1. Introducton Vector 3D model representaton n computer graphcs requres

More information

Conditional Speculative Decimal Addition*

Conditional Speculative Decimal Addition* Condtonal Speculatve Decmal Addton Alvaro Vazquez and Elsardo Antelo Dep. of Electronc and Computer Engneerng Unv. of Santago de Compostela, Span Ths work was supported n part by Xunta de Galca under grant

More information

An Efficient Garbage Collection for Flash Memory-Based Virtual Memory Systems

An Efficient Garbage Collection for Flash Memory-Based Virtual Memory Systems S. J and D. Shn: An Effcent Garbage Collecton for Flash Memory-Based Vrtual Memory Systems 2355 An Effcent Garbage Collecton for Flash Memory-Based Vrtual Memory Systems Seunggu J and Dongkun Shn, Member,

More information

Performance Study of Parallel Programming on Cloud Computing Environments Using MapReduce

Performance Study of Parallel Programming on Cloud Computing Environments Using MapReduce 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

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

Run-Time Operator State Spilling for Memory Intensive Long-Running Queries

Run-Time Operator State Spilling for Memory Intensive Long-Running Queries Run-Tme Operator State Spllng for Memory Intensve Long-Runnng Queres Bn Lu, Yal Zhu, and lke A. Rundenstener epartment of Computer Scence, Worcester Polytechnc Insttute Worcester, Massachusetts, USA {bnlu,

More information

A Hybrid Genetic Algorithm for Routing Optimization in IP Networks Utilizing Bandwidth and Delay Metrics

A Hybrid Genetic Algorithm for Routing Optimization in IP Networks Utilizing Bandwidth and Delay Metrics A Hybrd Genetc Algorthm for Routng Optmzaton n IP Networks Utlzng Bandwdth and Delay Metrcs Anton Redl Insttute of Communcaton Networks, Munch Unversty of Technology, Arcsstr. 21, 80290 Munch, Germany

More information

DESIGNING TRANSMISSION SCHEDULES FOR WIRELESS AD HOC NETWORKS TO MAXIMIZE NETWORK THROUGHPUT

DESIGNING TRANSMISSION SCHEDULES FOR WIRELESS AD HOC NETWORKS TO MAXIMIZE NETWORK THROUGHPUT DESIGNING TRANSMISSION SCHEDULES FOR WIRELESS AD HOC NETWORKS TO MAXIMIZE NETWORK THROUGHPUT Bran J. Wolf, Joseph L. Hammond, and Harlan B. Russell Dept. of Electrcal and Computer Engneerng, Clemson Unversty,

More information

USING GRAPHING SKILLS

USING GRAPHING SKILLS Name: BOLOGY: Date: _ Class: USNG GRAPHNG SKLLS NTRODUCTON: Recorded data can be plotted on a graph. A graph s a pctoral representaton of nformaton recorded n a data table. t s used to show a relatonshp

More information

ARTICLE IN PRESS. Signal Processing: Image Communication

ARTICLE IN PRESS. Signal Processing: Image Communication Sgnal Processng: Image Communcaton 23 (2008) 754 768 Contents lsts avalable at ScenceDrect Sgnal Processng: Image Communcaton journal homepage: www.elsever.com/locate/mage Dstrbuted meda rate allocaton

More information

Outline. Digital Systems. C.2: Gates, Truth Tables and Logic Equations. Truth Tables. Logic Gates 9/8/2011

Outline. Digital Systems. C.2: Gates, Truth Tables and Logic Equations. Truth Tables. Logic Gates 9/8/2011 9/8/2 2 Outlne Appendx C: The Bascs of Logc Desgn TDT4255 Computer Desgn Case Study: TDT4255 Communcaton Module Lecture 2 Magnus Jahre 3 4 Dgtal Systems C.2: Gates, Truth Tables and Logc Equatons All sgnals

More information

Proper Choice of Data Used for the Estimation of Datum Transformation Parameters

Proper Choice of Data Used for the Estimation of Datum Transformation Parameters Proper Choce of Data Used for the Estmaton of Datum Transformaton Parameters Hakan S. KUTOGLU, Turkey Key words: Coordnate systems; transformaton; estmaton, relablty. SUMMARY Advances n technologes and

More information

A RECONFIGURABLE ARCHITECTURE FOR MULTI-GIGABIT SPEED CONTENT-BASED ROUTING. James Moscola, Young H. Cho, John W. Lockwood

A RECONFIGURABLE ARCHITECTURE FOR MULTI-GIGABIT SPEED CONTENT-BASED ROUTING. James Moscola, Young H. Cho, John W. Lockwood A RECONFIGURABLE ARCHITECTURE FOR MULTI-GIGABIT SPEED CONTENT-BASED ROUTING James Moscola, Young H. Cho, John W. Lockwood Dept. of Computer Scence and Engneerng Washngton Unversty, St. Lous, MO {jmm5,

More information

The Data Warehouse in a Distributed Utility Environment

The Data Warehouse in a Distributed Utility Environment The Data Warehouse n a Dstrbuted Utlty Envronment Charles A. Mllgan Dstngushed Engneer, Sun Mcrosystems Charles.mllgan@sun.com Abstract Utlty provsonng, Grd resource management, nstant copy kosks, and

More information

Fibre-Optic AWG-based Real-Time Networks

Fibre-Optic AWG-based Real-Time Networks Fbre-Optc AWG-based Real-Tme Networks Krstna Kunert, Annette Böhm, Magnus Jonsson, School of Informaton Scence, Computer and Electrcal Engneerng, Halmstad Unversty {Magnus.Jonsson, Krstna.Kunert}@de.hh.se

More information

Advanced Computer Networks

Advanced Computer Networks Char of Network Archtectures and Servces Department of Informatcs Techncal Unversty of Munch Note: Durng the attendance check a stcker contanng a unque QR code wll be put on ths exam. Ths QR code contans

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

Chapter 1. Introduction

Chapter 1. Introduction Chapter 1 Introducton 1.1 Parallel Processng There s a contnual demand for greater computatonal speed from a computer system than s currently possble (.e. sequental systems). Areas need great computatonal

More information

Application of Clustering Algorithm in Big Data Sample Set Optimization

Application of Clustering Algorithm in Big Data Sample Set Optimization Applcaton of Clusterng Algorthm n Bg Data Sample Set Optmzaton Yutang Lu 1, Qn Zhang 2 1 Department of Basc Subjects, Henan Insttute of Technology, Xnxang 453002, Chna 2 School of Mathematcs and Informaton

More information

Routing in Degree-constrained FSO Mesh Networks

Routing in Degree-constrained FSO Mesh Networks Internatonal Journal of Hybrd Informaton Technology Vol., No., Aprl, 009 Routng n Degree-constraned FSO Mesh Networks Zpng Hu, Pramode Verma, and James Sluss Jr. School of Electrcal & Computer Engneerng

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

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

Parallel Data Processing in the Cloud using Nephele

Parallel Data Processing in the Cloud using Nephele Internatonal Journal of Computer Applcatons (0975 8887) Parallel Data Processng n the Cloud usng Nephele Mayura D. Tapkre PG Student M B E Socety s College of Engneerng, Ambajoga Maharastra, Inda B. M.

More information

Towards Autonomous Service Composition in A Grid Environment

Towards Autonomous Service Composition in A Grid Environment Towards Autonomous Servce Composton n A Grd Envronment Wllam K. Cheung +, Jmng Lu +, Kevn H. Tsang +, Raymond K. Wong ++ Department of Computer Scence + Hong Kong Baptst Unversty Hong Kong {wllam,jmng,hhtsang}@comp.hkbu.edu.hk

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

mquest Quickstart Version 11.0

mquest Quickstart Version 11.0 mquest Quckstart Verson 11.0 cluetec GmbH Emmy-Noether-Straße 17 76131 Karlsruhe Germany www.cluetec.de www.mquest.nfo cluetec GmbH Karlsruhe, 2016 Document verson 5 27.04.2016 16:59 > Propretary notce

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

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

QOS AWARE HW/SW PARTITIONING ON RUN-TIME RECONFIGURABLE MULTIMEDIA PLATFORMS

QOS AWARE HW/SW PARTITIONING ON RUN-TIME RECONFIGURABLE MULTIMEDIA PLATFORMS QOS AWARE HW/SW PARTITIONING ON RUN-TIE RECONFIGURABE UTIEDIA PATFORS N. Pham Ngoc, G. afrut 2, J-Y. gnolet 2, G. Deconnc, and R. auwerens 2 Katholee Unverstet euven-esat/eecta Kasteelpar Arenberg 0 B-300

More information

A fair buffer allocation scheme

A fair buffer allocation scheme A far buffer allocaton scheme Juha Henanen and Kalev Klkk Telecom Fnland P.O. Box 228, SF-330 Tampere, Fnland E-mal: juha.henanen@tele.f Abstract An approprate servce for data traffc n ATM networks requres

More information

A Semi-Distributed Load Balancing Architecture and Algorithm for Heterogeneous Wireless Networks

A Semi-Distributed Load Balancing Architecture and Algorithm for Heterogeneous Wireless Networks A Sem-Dstrbuted oad Balancng Archtecture and Algorthm for Heterogeneous reless Networks Md. Golam Rabul Ala Choong Seon Hong * Kyung Hee Unversty, Korea rob@networkng.khu.ac.kr, cshong@khu.ac.kr Abstract

More information

Energy Aware Virtual Machine Migration Techniques for Cloud Environment

Energy Aware Virtual Machine Migration Techniques for Cloud Environment Energy Aware rtual Machne Mgraton Technques for Cloud Envronment Kamal Gupta Department of CSE MMU, Sadopur jay Katyar, PhD Department of CSE MMU, Mullana ABSTRACT Cloud Computng offers ndspensable nfrastructure

More information

Intro. Iterators. 1. Access

Intro. Iterators. 1. Access Intro Ths mornng I d lke to talk a lttle bt about s and s. We wll start out wth smlartes and dfferences, then we wll see how to draw them n envronment dagrams, and we wll fnsh wth some examples. Happy

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

Towards High Fidelity Network Emulation

Towards High Fidelity Network Emulation Towards Hgh Fdelty Network Emulaton Lanje Cao, Xangyu Bu, Sona Fahmy, Syuan Cao Department of Computer Scence, Purdue nversty, West Lafayette, IN, SA E-mal: {cao62, xb, fahmy, cao208}@purdue.edu Abstract

More information

Sum of Linear and Fractional Multiobjective Programming Problem under Fuzzy Rules Constraints

Sum of Linear and Fractional Multiobjective Programming Problem under Fuzzy Rules Constraints Australan Journal of Basc and Appled Scences, 2(4): 1204-1208, 2008 ISSN 1991-8178 Sum of Lnear and Fractonal Multobjectve Programmng Problem under Fuzzy Rules Constrants 1 2 Sanjay Jan and Kalash Lachhwan

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