Harvesting idle CPU resources for desktop grid computing while limiting the slowdown generated to end-users

详细信息    查看全文

• 作者：Eduardo Rosales ; Germán Sotelo ; Antonio de la Vega ; César O. Díaz…
• 关键词：Desktop grid computing ; Volunteer computing ; Slowdown ; Overclocking ; BOINC ; UnaCloud
• 刊名：Cluster Computing
• 出版年：2015
• 出版时间：December 2015
• 年：2015
• 卷：18
• 期：4
• 页码：1331-1350
• 全文大小：4,292 KB
• 参考文献：1.Anderson, D.P.: Boinc: a system for public-resource computing and storage. In: Proceedings of the 5th IEEE/ACM International Workshop on Grid Computing, GRID -4, pp. 4-0. IEEE Computer Society, Washington, DC, USA (2004). doi:10.-109/?GRID.-004.-4
  2.Anderson, D.P.: Volunteer computing: the ultimate cloud. ACM Crossroads 16(3), 7-0 (2010)CrossRef
  3.Anderson, D.P., Fedak, G.: The computational and storage potential of volunteer computing. In: Sixth IEEE International Symposium on Cluster Computing and the Grid, 2006. CCGRID 06, vol. 1, pp. 73-0. (2006)
  4.Andrade, N., Cirne, W., Brasileiro, F., Roisenberg, P.: OurGrid: An Approach to Easily Assemble Grids with Equitable Resource Sharing. Lecture Notes in Computer Science, vol. 2862, pp. 61-6. Springer, Berlin (2003). doi:10.-007/-0968987-4
  5.Atlas, J., Estrada, T., Decker, K., Taufer, M.: Balancing Scientist Needs and Volunteer Preferences in Volunteer Computing Using Constraint Optimization. Lecture Notes in Computer Science, vol. 5544(15), pp. 143-52. Springer, Berlin (2009). doi:10.-007/-78-3-642-01970-8-15
  6.BOINC: Boinccmd tool—boinc. http://?boinc.?berkeley.?edu/?wiki/?Boinccmd-tool
  7.BOINC: Client scheduling policies—boinc. http://?boinc.?berkeley.?edu/?trac/?wiki/?ClientSched
  8.BOINC: Seti@home—boinc. http://?setiathome.?ssl.?berkeley.?edu/-/span> . Accessed 23 Nov 2014
  9.BOINC: Volunteer computing (2013). http://?boinc.?berkeley.?edu/?trac/?wiki/?VolunteerComputi?ng . Accessed 13 Oct 2014
  10.BOINC: Einstein@home. http://?boinc.?berkeley.?edu/?wiki/?Einstein@Home (2014)
  11.BOINC: Primegrid@home. http://?www.?primegrid.?com/-/span> (2014)
  12.Bunci, P., Aguado-Sanchez, C., Blomer, J., Franco, L., Harutyunian, A., Mato, P., Yao, Y.: Cernvm—a virtual software appliance for lhc applications. J. Phys. 219(4), 43-3 (2010). doi:10.-088/-742-6596/-19/-/-42003
  13.Castro, H., Rosales, E., Villamizar, M., Jiménez, A.: Unagrid: on demand opportunistic desktop grid. In: Proceedings of the 2010 10th IEEE/ACM International Conference on Cluster, Cloud and Grid Computing, CCGRID -0, pp. 661-66. IEEE Computer Society, Washington, DC, USA (2010). doi:10.-109/?CCGRID.-010.-9
  14.Charles, J., Jassi, P., Ananth, N.S., Sadat, A., Fedorova, A.: Evaluation of the intel core i7 turbo boost feature. In: IEEE International Symposium on Workload Characterization, 2009. IISWC 2009, pp. 188-97 (2009). doi:10.-109/?IISWC.-009.-306782
  15.Corporation, I.: First the tick, now the tock: Next generation intel microarchitecture (nehalem). Report (2009)
  16.Corporation, I.: Enhanced intel speedstep technology—how to document. Report (2015). http://?www.?intel.?com/?cd/?channel/?reseller/?asmo-na/?eng/-03838.?h
  17.Diaz, C., Pecero, J., Bouvry, P., Sotelo, G., Villamizar, M., Castro, H.: Performance evaluation of an iaas opportunistic cloud computing. In: 14th IEEE/ACM International Symposium on Cluster, Cloud and Grid Computing (CCGrid), 2014, pp. 546-47 (2014). doi:10.-109/?CCGrid.-014.-16
  18.Domingues, P., Araujo, F., Silva, L.: Evaluating the performance and intrusiveness of virtual machines for desktop grid computing. In: IEEE International Symposium on Parallel & Distributed Processing, 2009. IPDPS 2009, pp. 1- (2009). doi:10.-109/?IPDPS.-009.-161134
  19.Foster, I., Iamnitchi, A.: On Death, Taxes, and the Convergence of Peer-to-Peer and Grid Computing. Lecture Notes in Computer Science, vol. 2735(11), pp. 118-28. Springer, Berlin (2003). doi:10.-007/-78-3-540-45172-3-11
  20.Hyperic, I.: Sigar api. http://?www.?hyperic.?com/?products/?sigar
  21.Intel: Intel math kernel library linpack download. https://?software.?intel.?com/?en-us/?articles/?intel-math-kernel-l
  22.Intel: Intel power gadget. https://?software.?intel.?com/?en-us/?articles/?intel-power-gadget-20
  23.Intel: Processors. http://?www.?intel.?com/?support/?processors/?corei7/?sb/?cs-032279.?h
  24.Intel: Processors. http://?www.?intel.?com/?support/?processors/?corei5/?sb/?CS-032278.?h
  25.Intel: Intel turbo boost technology in intel core microarchitecture (nehalem) based processors. Report (2008). http://?files.?shareholder.?com/-/span>
  26.Kondo, D., Fedak, G., Cappello, F., Chien, A.A., Casanova, H.: Characterizing resource availability in enterprise desktop grids. Future Gener. Comput. Syst. 23(7), 888-03 (2007). doi:10.-016/?j.?future.-006.-1.-01 CrossRef
  27.Laadan, O., Nieh, J.: Operating system virtualization: practice and experience (2010). doi:10.-145/-815695.-815717
  28.Lo, D., Kozyrakis, C.: Dynamic management of turbomode in modern multi-core chips. In: IEEE 20th International Symposium on High Performance Computer Architecture (HPCA), 2014, pp. 603-13 (2014). doi:10.-109/?HPCA.-014.-835969
  29.Marosi, A., Kovnács, J., Kacsuk, P.: Towards a volunteer cloud system. Future Gener. Comput. Syst. 29(6), 1442-451 (2013). doi:10.-016/?j.?future.-012.-3.-13 CrossRef
  3
• 作者单位：Eduardo Rosales (1)
  Germán Sotelo (1)
  Antonio de la Vega (1)
  César O. Díaz (1)
  Carlos E. Gómez (1) (2)
  Harold Castro (1)
  
  1. Systems and Computing Engineering Department, School of Engineering, Universidad de los Andes, Bogotá, Colombia
  2. Universidad del Quindio, Armenia, Colombia
  
• 刊物类别：Computer Science
• 刊物主题：Processor Architectures
  Operating Systems
  Computer Communication Networks
  
• 出版者：Springer Netherlands
• ISSN：1573-7543

文摘

We address the challenge of both harvesting idle CPU resources on off-the-shelf desktops donated to Desktop Grid Computing while at once limiting the slowdown generated to the resource owner, also known as end-user, to customized values. In this context, slowdown is studied as the increase in completion times of end-user tasks while a Desktop Grid harvests idle CPU resources by executing CPU intensive workloads. To achieve this, we deploy two Desktop Grids, one virtualization-based (UnaCloud) and one agent-based (BOINC). We then quantify the slowdown generated to simultaneously-running, end-user tasks. The results show that dynamic performance and energy-efficient technologies, specifically overclocking features, directly affect the slowdown generated to the end-user when incorporated into the processor used by the Desktop Grid. Furthermore, we propose, implement, and test a first set of resource allocation policies for the BOINC client in order to effectively harvest idle CPU resources while avoiding to exceed a customizable slowdown limit. Keywords Desktop grid computing Volunteer computing Slowdown Overclocking BOINC UnaCloud