Abstract
Amazon EC2 and Google Compute Engine (GCE) have recently introduced a new class of virtual machines called "burstable" instances that are cheaper than even the smallest traditional/regular instances. These lower prices come with reduced average capacity and increased variance. Using measurements from both EC2 and GCE, we identify key idiosyncrasies of resource capacity dynamism for burstable instances that set them apart from other instance types. Most importantly, certain resources for these instances appear to be regulated by deterministic token bucket like mechanisms. We find widely different types of disclosures by providers of the parameters governing these regulation mechanisms: full disclosure (e.g., CPU capacity for EC2 t2 instances), partial disclosure (e.g., CPU capacity and remote disk IO bandwidth for GCE shared-core instances), or no disclosure (network bandwidth for EC2 t2 instances). A tenant modeling these variations as random phenomena (as some recent work suggests) might make sub-optimal procurement and operation decisions. We present modeling techniques for a tenant to infer the properties of these regulation mechanisms via simple offline measurements. We also present two case studies of how certain memcached workloads might benefit from our modeling when operating on EC2 by: (i) augmenting cheap but low availability in-memory storage offered by spot instances with backup of popular content on burstable instances, and (ii) temporal multiplexing of multiple burstable instances to achieve the CPU or network bandwidth (and thereby throughput) equivalent of a more expensive regular EC2 instance.
- B. Adler. Amazon t2 use cases. http://www.rightscale.com/blog/cloud-cost-analysis/will-aws-t2-replace-30-percent-instances-not-so-fast.Google Scholar
- Amazon EC2 EBS Volume. http://docs.aws.amazon.com/AWSEC2/latest/UserGuide/EBSVolumeTypes.html.Google Scholar
- Amazon Elastic Block Store (EBS). https://aws.amazon.com/ebs/.Google Scholar
- Amazon instance types. https://aws.amazon.com/ec2/instance-types/.Google Scholar
- Amazon spot instances. https://aws.amazon.com/ec2/spot/, 2016.Google Scholar
- Amazon EC2 FAQs: What is an EC2 Compute Unit and why did you introduce it? https://aws.amazon.com/ec2/faqs/#hardware-information.Google Scholar
- Amazon cloudwatch - cloud and network monitoring services. https://aws.amazon.com/cloudwatch/.Google Scholar
- S. K. Barker and P. Shenoy. Empirical evaluation of latency-sensitive application performance in the cloud. In Proc. ACM Multimedia, 2010. Google ScholarDigital Library
- N. Budhiraja, K. Marzullo, F. B. Schneider, and S. Toueg. The primary-backup approach. In Distributed Systems (2nd Ed.), pages 199--216. ACM Press/Addison-Wesley Publishing Co., 1993. Google ScholarDigital Library
- Amazon burstable instances. http://docs.aws.amazon.com/AWSEC2/latest/UserGuide/t2-instances.html.Google Scholar
- E. D. Carreno, M. Diener, E. Cruz, and P. Navaux. Automatic communication optimization of parallel applications in public clouds. In Proc. IEEE/ACM CCGRID, 2016.Google ScholarDigital Library
- CloudLook. http://www.cloudlook.com.Google Scholar
- C. Delimitrou and C. Kozyrakis. Hcloud: Resource-efficient provisioning in shared cloud systems. In Proc. ACM ASPLOS, 2016. Google ScholarDigital Library
- EC2 variability: The numbers revealed. http://tech.mangot.com/roller/dave/entry/ec2_variability_the_numbers_revealed.Google Scholar
- Y. El-Khamra, H. Kim, S. Jha, and M. Parashar. Exploring the performance fluctuations of HPC workloads on clouds. In Proc. IEEE CloudCom, 2010. Google ScholarDigital Library
- B. Farley, A. Juels, V. Varadarajan, T. Ristenpart, K. D. Bowers, and M. M. Swift. More for your money: exploiting performance heterogeneity in public clouds. In Proc. ACM SOCC, 2012. Google ScholarDigital Library
- A. Gandhi and J. Chan. Analyzing the Network for AWS Distributed Cloud Computing. Proc. ACM SIGMETRICS PER, 2015. Google ScholarDigital Library
- GCE Disk Performance. https://cloud.google.com/compute/docs/disks/performance.Google Scholar
- M. Hajjat, R. Liu, Y. Chang, T. E. Ng, and S. Rao. Application-specific configuration selection in the cloud: impact of provider policy and potential of systematic testing. In Proc. IEEE INFOCOM, 2015.Google ScholarCross Ref
- J. Heinanen, T. Finland, and R. Guerin. A two rate three color marker. RFC 2698 available at www.ietf.org, 1999. Google ScholarDigital Library
- A. Iosup, N. Yigitbasi, and D. Epema. On the performance variability of production cloud services. In Proc. of IEEE/ACM CCGrid, 2011. Google ScholarDigital Library
- iPerf. https://iperf.fr/iperf-download.php.Google Scholar
- A. Kopytov. Sysbench manual. MySQL AB, 2012.Google Scholar
- P. Leitner and J. Cito. Patterns in the Chaos - A Study of Performance Variation and Predictability in Public IaaS Clouds. ACM Transactions on Internet Technology (TOIT), 2016. Google ScholarDigital Library
- P. Leitner and J. Scheuner. Bursting with Possibilities--An Empirical Study of Credit-Based Bursting Cloud Instance Types. In Proc. IEEE/ACM UCC, 2015.Google Scholar
- A. Li, X. Yang, S. Kandula, and M. Zhang. CloudCmp: comparing public cloud providers. In Proc. ACM SIGCOMM IMC, 2010. Google ScholarDigital Library
- lookbusy - a synthetic load generator. https://www.devin.com/lookbusy/.Google Scholar
- GCE Machine Types. https://cloud.google.com/compute/docs/machine-types.Google Scholar
- Memcached. http://memcached.org/.Google Scholar
- A. Nhem. Cloudability. https://blog.cloudability.com/how-cost-efficient-is-the-new-burstable-aws-t2-large/, 2016.Google Scholar
- S. Ostermann, A. Iosup, N. Yigitbasi, R. Prodan, T. Fahringer, and D. Epema. A performance analysis of EC2 cloud computing services for scientific computing. Cloud computing, 2009.Google Scholar
- Z. Ou, H. Zhuang, J. K. Nurminen, A. Yl\"a-J\"a\"aski, and P. Hui. Exploiting hardware heterogeneity within the same instance type of Amazon EC2. In Proc. USENIX HotCloud, 2012. Google ScholarDigital Library
- J. Read. Cloudharmony. http://blog.cloudharmony.com/2010/05/what-is-ecu-cpu-benchmarking-in-cloud.html.Google Scholar
- M. S. Rehman and M. F. Sakr. Initial findings for provisioning variation in cloud computing. In Proc. IEEE CloudCom, 2010. Google ScholarDigital Library
- C. Reiss, A. Tumanov, G. R. Ganger, R. H. Katz, and M. A. Kozuch. Heterogeneity and dynamicity of clouds at scale: Google trace analysis. In Proc. ACM SOCC, 2012. Google ScholarDigital Library
- J. Schad, J. Dittrich, and J. Quiané-Ruiz. Runtime measurements in the cloud: observing, analyzing, and reducing variance. Proc. VLDB Endowment, 2010. Google ScholarDigital Library
- GCE Micro History. https://en.scio.pw/Google_Compute_Engine.Google Scholar
- GCE Preemptible Instances. https://cloud.google.com/compute/docs/instances/preemptible.Google Scholar
- GCE Shared-Core Instances. https://cloud.google.com/compute/docs/machine-types#sharedcore.Google Scholar
- A. Singh, J. Ong, A. Agarwal, G. Anderson, A. Armistead, R. Bannon, S. Boving, G. Desai, B. Felderman, P. Germano, et al. Jupiter rising: A decade of Clos topologies and centralized control in Google's datacenter network. ACM SIGCOMM Computer Communication Review, 2015. Google ScholarDigital Library
- STREAM, 2016. http://www.cs.virginia.edu/stream/.Google Scholar
- EC2 t2 Instances History. https://aws.amazon.com/about-aws/whats-new/2014/07/01/introducing-t2-the-new-low-cost-general-purpose/instance-type-for-amazon-ec2/, 2014.Google Scholar
- Linux traffic control: Token bucket filter. http://lartc.org/manpages/tc-tbf.html.Google Scholar
- P. B. Teregowda and C. L. Giles. Scaling seersuite in the cloud. In Proc. IEEE Int'l Conf. on Cloud Engr, 2013. Google ScholarDigital Library
- B. Urgaonkar, P. J. Shenoy, and T. Roscoe. Resource overbooking and application profiling in shared hosting platforms. In Proc. USENIX OSDI, 2002. Google ScholarDigital Library
- C. Wang, Q. Liang, and B. Urgaonkar. An empirical analysis of amazon ec2 spot instance features affecting cost-effective resource procurement. In Proc. ACM/SPEC ICPE, 2017. Google ScholarDigital Library
- C. Wang, B. Urgaonkar, A. Gupta, L. Chen, R. Birke, and G. Kesidis. Effective capacity modulation as an explicit control knob for public cloud profitability. In Proc. IEEE ICAC, 2016.Google ScholarCross Ref
- G. Wang and T. E. Ng. The impact of virtualization on network performance of amazon EC2 data center. In Proc. IEEE INFOCOM, 2010. Google ScholarDigital Library
- J. Wen, L. Lu, G. Casale, and E. Smirni. Less can be more: Micro-managing VMs in Amazon EC2. In Proc. IEEE CLOUD, 2015. Google ScholarDigital Library
- Z. Xu, C. Stewart, N. Deng, and X. Wang. Blending on-demand and spot instances to lower costs for in-memory storage. In Proc. IEEE INFOCOM, 2016.Google ScholarCross Ref
- Yahoo Cloud Serving Benchmark (YCSB. https://research.yahoo.com/news/yahoo-cloud-serving-benchmark.Google Scholar
- X. Zhang, E. Tune, R. Hagmann, R. Jnagal, V. Gokhale, and J. Wilkes. CP$\mboxI^2$: CPU performance isolation for shared compute clusters. In Proc. Eurosys, 2013. Google ScholarDigital Library
- Z. Zhang, L. Cherkasova, and B. T. Loo. Exploiting cloud heterogeneity for optimized cost/performance mapreduce processing. In Proc. International Workshop on Cloud Data and Platforms, 2014. Google ScholarDigital Library
- Z. Zhang, L. Cherkasova, and B. T. Loo. Optimizing cost and performance trade-offs for mapreduce job processing in the cloud. In Proc. IEEE NOMS, 2014.Google ScholarCross Ref
Index Terms
- Using Burstable Instances in the Public Cloud: Why, When and How?
Recommendations
Exploiting Spot and Burstable Instances for Improving the Cost-efficacy of In-Memory Caches on the Public Cloud
EuroSys '17: Proceedings of the Twelfth European Conference on Computer SystemsIn order to keep the costs of operating in-memory storage on the public cloud low, we devise novel ideas and enabling modeling and optimization techniques for combining conventional Amazon EC2 instances with the cheaper spot and burstable instances. ...
Using Burstable Instances in the Public Cloud: Why, When and How?
SIGMETRICS '17 Abstracts: Proceedings of the 2017 ACM SIGMETRICS / International Conference on Measurement and Modeling of Computer SystemsTo attract more customers, public cloud providers offer virtual machine (instance) types that trade off lower prices for poorer capacities. As one salient approach, the providers employ aggressive statistical multiplexing of multiple cheaper instances ...
Using Burstable Instances in the Public Cloud: Why, When and How?
Performance evaluation reviewTo attract more customers, public cloud providers offer virtual machine (instance) types that trade off lower prices for poorer capacities. As one salient approach, the providers employ aggressive statistical multiplexing of multiple cheaper instances ...
Comments