Abstract
Large-scale datacenters (DCs) host tens of thousands of diverse applications each day. However, interference between colocated workloads and the difficulty of matching applications to one of the many hardware platforms available can degrade performance, violating the quality of service (QoS) guarantees that many cloud workloads require. While previous work has identified the impact of heterogeneity and interference, existing solutions are computationally intensive, cannot be applied online, and do not scale beyond a few applications.
We present Paragon, an online and scalable DC scheduler that is heterogeneity- and interference-aware. Paragon is derived from robust analytical methods, and instead of profiling each application in detail, it leverages information the system already has about applications it has previously seen. It uses collaborative filtering techniques to quickly and accurately classify an unknown incoming workload with respect to heterogeneity and interference in multiple shared resources. It does so by identifying similarities to previously scheduled applications. The classification allows Paragon to greedily schedule applications in a manner that minimizes interference and maximizes server utilization. After the initial application placement, Paragon monitors application behavior and adjusts the scheduling decisions at runtime to avoid performance degradations. Additionally, we design ARQ, a multiclass admission control protocol that constrains application waiting time. ARQ queues applications in separate classes based on the type of resources they need and avoids long queueing delays for easy-to-satisfy workloads in highly-loaded scenarios. Paragon scales to tens of thousands of servers and applications with marginal scheduling overheads in terms of time or state.
We evaluate Paragon with a wide range of workload scenarios, on both small and large-scale systems, including 1,000 servers on EC2. For a 2,500-workload scenario, Paragon enforces performance guarantees for 91% of applications, while significantly improving utilization. In comparison, heterogeneity-oblivious, interference-oblivious, and least-loaded schedulers only provide similar guarantees for 14%, 11%, and 3% of workloads. The differences are more striking in oversubscribed scenarios where resource efficiency is more critical.
- Alameldeen, A. R. and Wood, D. A. 2006. IPC considered harmful for multiprocessor workloads. IEEE Micro (Special Issue on Computer Architecture Simulation and Modeling). Google ScholarDigital Library
- Amazon EC2. http://aws.amazon.com/ec2/.Google Scholar
- Banga, G., Druschel, P., and Mogul, J. C. 1999. Resource containers: A new facility for resource management in server systems. In Proceedings of the Third Symposium on Operating Systems Design and Implementation (OSDI). Google ScholarDigital Library
- Barroso, L. 2011. Warehouse-scale computing: entering the teenage decade. In Proceedings of ISCA. Google ScholarCross Ref
- Barroso, L. and Hoelzle, U. 2009. The Datacenter as a Computer: An Introduction to the Design of Warehouse-Scale Machines. Morgan and Claypool. Google ScholarDigital Library
- Bell, R. M., Koren, Y., and Volinsky, C. 2007. The BellKor 2008 solution to the Netflix Prize. Tech. rep., AT&T Labs.Google Scholar
- Bertsimas, D., Gamarnik, D., and Tsitsiklis, J. N. 2001. Performance of multiclass Markovian queueing networks via piecewise linear Lyapunov functions. Ann. Appl. Probab. 11, 1384--1428.Google ScholarCross Ref
- Bienia, C., Kumar, S., Singh, J. P., and Li, K. 2008. The PARSEC benchmark suite: Characterization and architectural implications. In Proceedings of the 17th International Conference on Parallel Architectures and Compilation Techniques (PACT). Google ScholarDigital Library
- Bottou, L. 2010. Large-scale machine learning with stochastic gradient descent. In Proceedings of the International Conference on Computational Statistics (COMPSTAT).Google ScholarCross Ref
- Calder, B., Wang, J., Ogus, A., Nilakantan, N., Skjolsvold, A., McKelvie, S., Xu, Y., Srivastav, S., Wu, J., Simitci, H., Haridas, J., Uddaraju, C., Khatri, H., Edwards, A., Bedekar, V., Mainali, S., Abbasi, R., Agarwal, A., ul Haq, M. F., ul Haq, M. I., Bhardwaj, D., Dayanand, S., Adusumilli, A., McNett, M., Sankaran, S., Manivannan, K., and Rigas, L. 2011. Windows Azure storage: A highly available cloud storage service with strong consistency. In Proceedings of the 23rd ACM Symposium on Operating Systems Principles (SOSP). Google ScholarDigital Library
- Chase, J., Anderson, D., Thakar, P., Vahdat, A., and Doyle, R. 2001. Managing energy and server resources in hosting centers. In Proceedings of the 18th ACM Symposium on Operating Systems Principles (SOSP). Google ScholarDigital Library
- Craeynest, K. V., Jaleel, A., Eeckhout, L., Narvaez, P., and Emer, J. 2012. Scheduling heterogeneous multi-cores through performance impact estimation (PIE). In Proceedings of the International Symposium on Computer Architecture (ISCA). Google ScholarDigital Library
- Dai, J. G. 1995. On positive Harris recurrence of multiclass queueing networks: A unified approach via fluid limit models. Ann. Appl. Probab. 5, 49--77.Google ScholarCross Ref
- Dai, J. G. 1996. A fluid-limit model criterion for instability of multiclass queueing networks. Ann. Appl. Probab. 6, 751--757.Google ScholarCross Ref
- Delimitrou, C. and Kozyrakis, C. 2013a. iBench: Quantifying interference for datacenter applications. In Proceedings of the IEEE International Symposium on Workload Characterization (IISWC).Google Scholar
- Delimitrou, C. and Kozyrakis, C. 2013b. Paragon: QoS-aware scheduling for heterogeneous datacenters. In Proceedings of the 18th International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS). Google ScholarDigital Library
- Delimitrou, C. and Kozyrakis, C. 2013c. The Netflix challenge: Datacenter edition. IEEE Comput. Archit. Lett. (June). Google ScholarDigital Library
- Fedorova, A., Seltzer, M., and Smith, M. D. 2007. Improving performance isolation on chip multiprocessors via an operating system scheduler. In Proceedings of the 16th International Conference on Parallel Architecture and Compilation Techniques (PACT). Google ScholarDigital Library
- Gamarnik, D. 2000. On deciding stability of scheduling policies in queuing systems. In Proceedings of the 11th Annual ACM-SIAM Symposium on Discrete Algorithms. 467--476. Google ScholarDigital Library
- Google Compute Engine GCE. http://cloud.google.com/products/compute-engine.html.Google Scholar
- Ghodsi, A., Zaharia, M., Hindman, B., Konwinski, A., Shenker, S., and Stoica, I. 2011. Dominant resource fairness: Fair allocation of multiple resource types. In Proceedings of the 8th USENIX Conference on Networked Systems Design and Implementation (NSDI). Google ScholarDigital Library
- Gmach, D., Rolia, J., Cherkasova, L., and Kemper, A. 2007. Workload analysis and demand prediction of enterprise data center applications. In Proceedings of the 10th IEEE International Symposium on Workload Characterization (IISWC). Google ScholarDigital Library
- Govindan, S., Liu, J., Kansal, A., and Sivasubramaniam, A. 2011. Cuanta: Quantifying effects of shared on-chip resource interference for consolidated virtual machines. In Proceedings of the 2nd ACM Symposium on Cloud Computing (SOCC). Google ScholarDigital Library
- Hamilton, J. 2009. Internet-scale service infrastructure efficiency. In Proceedings of the 37th International Symposium on Computer Architecture (ISCA). Google ScholarDigital Library
- Hamilton, J. 2010. Cost of power in large-scale data centers. http://perspectives.mvdirona.com.Google Scholar
- Hasenbein, J. J. 1998. Stability, capacity, and scheduling of multiclass queuing networks. Ph.D. dissertation, Georgia Institute of Technology. Google ScholarDigital Library
- Hindman, B., Konwinski, A., Zaharia, M., Ghodsi, A., Joseph, A. D., Katz, R., Shenker, S., and Stoica, I. 2011. Mesos: A platform for fine-grained resource sharing in the data center. In Proceedings of the 8th USENIX Symposium on Networked Systems Design and Implementation (NSDI). Google ScholarDigital Library
- Jaleel, A., Mattina, M., and Jacob, B. L. 2006. Last level cache (LLC) performance of data mining workloads on a CMP—A case study of parallel bioinformatics workloads. In Proceedings of the 12th International Symposium on High-Performance Computer Architecture (HPCA-12).Google Scholar
- Katz, J. and Lindell, Y. 2007. Introduction to Modern Cryptography. Chapman & Hall/CRC Press. Google ScholarDigital Library
- Kiwiel, K. C. 2001. Convergence and efficiency of subgradient methods for quasiconvex minimization. Math. Program. (Series A), 90, 1, 1--25.Google ScholarCross Ref
- Kozyrakis, C., Kansal, A., Sankar, S., and Vaid, K. 2010. Server engineering insights for large-scale online services. IEEE Micro 30, 4, 8--19. DOI:http://dx.doi.org/10.1109/MM.2010.73. Google ScholarDigital Library
- Leverich, J. and Kozyrakis, C. 2010. On the energy (in)efficiency of Hadoop clusters. SIGOPS Oper. Syst. Rev. 44, 1, 61--65. Google ScholarDigital Library
- Lin, J. and Kolcz, A. 2012. Large-scale machine learning at Twitter. In Proceedings of the ACM SIGMOD Conference. Google ScholarDigital Library
- Mars, J. and Tang, L. 2013. Whare-map: heterogeneity in “homogeneous” warehouse-scale computers. In Proceedings of the 40th Annual International Symposium on Computer Architecture (ISCA). Google ScholarDigital Library
- Mars, J., Tang, L., and Hundt, R. 2011. Heterogeneity in “homogeneous”; warehouse-scale computers: A performance opportunity. IEEE Comput. Archit. Lett. 10, 2, 29--32. DOI:http://dx.doi.org/10.1109/L-CA.2011.14. Google ScholarDigital Library
- Meisner, D., Sadler, C. M., Barroso, L. A., Weber, W.-D., and Wenisch, T. F. 2011. Power management of online data-intensive services. In Proceedings of the 38th Annual International Symposium on Computer Architecture (ISCA). Google ScholarDigital Library
- Miller, B. L. 1969. A queuing reward system with several customer classes. Manage. Sci. 16, 3, 234--245.Google Scholar
- Narayanan, R., Ozisikyilmaz, B., Zambreno, J., Memik, G., and Choudhary, A. N. 2006. MineBench: A benchmark suite for data mining workloads. In Proceedings of the 9th IEEE International Symposium on Workload Characterization (IISWC).Google Scholar
- Nathuji, R., Isci, C., and Gorbatov, E. 2007. Exploiting platform heterogeneity for power efficient data centers. In Proceedings of the International Conference on Autonomic Computing (ICAC). Google ScholarDigital Library
- Nathuji, R., Kansal, A., and Ghaffarkhah, A. 2010. Q-Clouds: Managing performance interference effects for QoS-aware clouds. In Proceedings of the European Conference on Computer Systems (EuroSys'10). Google ScholarDigital Library
- Novakovi&cgrave;, D., Vasi&cgrave;, N., Novakovi&cgrave;, S., Kosti&cgrave;, D., and Bianchini, R. 2013. DeepDive: Transparently identifying and managing performance interference in virtualized environments. In Proceedings of the USENIX Annual Technical Conference (ATC). Google ScholarDigital Library
- Rackspace. Open Cloud. http://www.rackspace.com/.Google Scholar
- Rajaraman, A. and Ullman, J. 2011. Textbook on Mining of Massive Datasets. Rightscale. https://aws.amazon.com/solution-providers/isv/rightscale.Google Scholar
- Sanchez, D. and Kozyrakis, C. 2011. Vantage: Scalable and efficient fine-grain cache partitioning. In Proceedings of the 38th Annual International Symposium in Computer Architecture (ISCA-38). Google ScholarDigital Library
- Schein, A., Popescul, A., Ungar, L., and Pennock, D. 2002. Methods and metrics for cold-start recommendations. In Proceedings of the 25th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR). Google ScholarDigital Library
- Schwarzkopf, M., Konwinski, A., Abd-El-Malek, M., and Wilkes, J. 2013. Omega: Flexible, scalable schedulers for large compute clusters. In Proceedings of the 8th ACM European Conference on Computer Systems (EuroSys'13). Google ScholarDigital Library
- Shelepov, D., Alcaide, J. C. S., Jeffery, S., Fedorova, A., Perez, N., Huang, Z. F., Blagodurov, S., and Kumar, V. 2009. HASS: A scheduler for heterogeneous multicore systems. SIGOPS Oper. Syst. Rev. 43, 2. Google ScholarDigital Library
- Shen, Z., Subbiah, S., Gu, X., and Wilkes, J. 2011. CloudScale: elastic resource scaling for multi-tenant cloud systems. In Proceedings of the 2nd ACM Symposium on Cloud Computing (SOCC). Google ScholarDigital Library
- Sun, J., Xie, Y., Zhang, H., and Faloutsos, C. 2008. Less is more: Compact matrix decomposition for large sparse graphs. J. Stat. Anal. Data Mining 1, 1.Google ScholarCross Ref
- Tanenbaum, A. S. 2007. Modern Operating Systems. 3rd Ed. Peason Education, Inc. Google ScholarDigital Library
- Vasić, N., Novaković, D., Miučin, S., Kostić, D., and Bianchini, R. 2012. Deja vu: accelerating resource allocation in virtualized environments. In Proceedings of the 17th International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS). Google ScholarDigital Library
- vMotion. Migrate VMs with Zero Downtime. http//www.vmware.com/products/vmotion.Google Scholar
- VMWare-DRS. 2012. Distributed resource scheduler: design, implementation and lessons learned. VMware Tech. J. 1, 1.Google Scholar
- VMWare vSphere. http://www.vmware.com/products/vsphere/.Google Scholar
- Weng, L.-T., Yue, X., Yuefeng, L., and Nayak, R. 2008. Exploiting item taxonomy for solving cold-start problem in recommendation making. In Proceedings of the 20th IEEE International Conference on Tools with Artificial Intelligence (ICTAI). Google ScholarDigital Library
- Wenisch, T. F., Wunderlich, R. E., Ferdman, M., Ailamaki, A., Falsafi, B., and Hoe, J. C. 2006. SimFlex: Statistical sampling of computer system simulation. IEEE MICRO 26, 4. Google ScholarDigital Library
- Windows Azure. http://www.windowsazure.com/.Google Scholar
- Witten, I. H., Frank, E., and Holmes, G. 2011. Data Mining: Practical Machine Learning Tools and Techniques. 3rd Ed. Google ScholarDigital Library
- Woo, S. C., Ohara, M., Torrie, E., Singh, J. P., and Gupta, A. 1995. The SPLASH-2 programs: Characterization and methodological considerations. In Proceedings of the 22nd International Symposium on Computer Architecture (ISCA). Google ScholarDigital Library
- Xenserver. 6.1. http://www.citrix.com/xenserver/.Google Scholar
- Yang, H., Breslow, A., Mars, J., and Tang, L. 2013. Bubble-flux: Precise online QoS management for increased utilization in warehouse scale computers. In Proceedings of the 40th Annual International Symposium on Computer Architecture (ISCA). Google ScholarDigital Library
- Zaharia, M., Chowdhury, M., Das, T., Dave, A., Ma, J., McCauley, M., Franklin, M. J., Shenker, S., and Stoica, I. 2012. Spark: Cluster computing with working sets. In Proceedings of the 9th USENIX Symposium on Networked Systems Design and Implementation (NSDI).Google Scholar
- Zhang, X., Tune, E., Hagmann, R., Jnagal, R., Gokhale, V., and Wilkes, J. 2013. CPI2: CPU performance isolation for shared compute clusters. In Proceedings of the 8th ACM European Conference on Computer Systems (EuroSys'13). Google ScholarDigital Library
- Zhang, Z.-K., Liu, C., Zhang, Y.-C., and Zhou, T. 2010. Solving the cold-start problem in recommender systems with social tags. arXiv:1004.3732v2.Google Scholar
- Zhu, X., Young, D., Watson, B. J., Wang, Z., Rolia, J., Singhal, S., Mckee, B., Hyser, C., Gmach, D., Gardner, R., Christian, T., and Cherkasova, L. 2009. 1000 Islands: An integrated approach to resource management for virtualized datacenters. J. Cluster Comput. 12, 1. Google ScholarDigital Library
Index Terms
- QoS-Aware scheduling in heterogeneous datacenters with paragon
Recommendations
Paragon: QoS-aware scheduling for heterogeneous datacenters
ASPLOS '13: Proceedings of the eighteenth international conference on Architectural support for programming languages and operating systemsLarge-scale datacenters (DCs) host tens of thousands of diverse applications each day. However, interference between colocated workloads and the difficulty to match applications to one of the many hardware platforms available can degrade performance, ...
Paragon: QoS-aware scheduling for heterogeneous datacenters
ASPLOS '13Large-scale datacenters (DCs) host tens of thousands of diverse applications each day. However, interference between colocated workloads and the difficulty to match applications to one of the many hardware platforms available can degrade performance, ...
Paragon: QoS-aware scheduling for heterogeneous datacenters
ASPLOS '13Large-scale datacenters (DCs) host tens of thousands of diverse applications each day. However, interference between colocated workloads and the difficulty to match applications to one of the many hardware platforms available can degrade performance, ...
Comments