Burstable Instances for Clouds: Performance Modeling, Equilibrium Analysis, and Revenue Maximization

Leading cloud providers recently introduced a new instance type named burstable instances to better match the time-varying workloads of tenants and further reduce their costs. In the research community, however, little has been done to understand burstable instances from a theoretical perspective. This paper presents the first unified framework to model, analyze, and optimize the operation of burstable instances. Specifically, we model the resource provisioning of burstable instances, identify key performance metrics, and derive the analytical performance given the resource provisioning decisions. We then characterize the equilibrium behind tenants’ responses to the prices offered for different burstable instance service classes, taking into account the impact of tenants’ actions on the performance achieved by each service class. In addition, we investigate how a cloud provider can leverage knowledge of this equilibrium to find the prices that maximize its total revenue. Finally, we validate our framework on real traces and demonstrate its usage to price burstable offerings in a public cloud.

[1]  Sangtae Ha,et al.  Offering Supplementary Network Technologies: Adoption Behavior and Offloading Benefits , 2015, IEEE/ACM Transactions on Networking.

[2]  Ricardo Bianchini,et al.  Resource Central: Understanding and Predicting Workloads for Improved Resource Management in Large Cloud Platforms , 2017, SOSP.

[3]  Qingsheng Zhu,et al.  Percentile Performance Estimation of Unreliable IaaS Clouds and Their Cost-Optimal Capacity Decision , 2017, IEEE Access.

[4]  R. Srikant,et al.  The Price of Simplicity , 2007, 2007 Conference Record of the Forty-First Asilomar Conference on Signals, Systems and Computers.

[5]  Liang Zheng,et al.  How to Bid the Cloud , 2015, Comput. Commun. Rev..

[6]  Philipp Leitner,et al.  Bursting with Possibilities -- An Empirical Study of Credit-Based Bursting Cloud Instance Types , 2015, 2015 IEEE/ACM 8th International Conference on Utility and Cloud Computing (UCC).

[7]  Eric Bouillet,et al.  Efficient resource provisioning in compute clouds via VM multiplexing , 2010, ICAC '10.

[8]  Erik Elmroth,et al.  Incentivizing self-capping to increase cloud utilization , 2017, SoCC.

[9]  Danny H. K. Tsang,et al.  On Power-Peak-Aware Scheduling for Large-Scale Shared Clusters , 2020, IEEE Transactions on Big Data.

[10]  Robert Birke,et al.  Optimizing for Tail Sojourn Times of Cloud Clusters , 2018, IEEE Transactions on Cloud Computing.

[11]  Baochun Li,et al.  Dynamic Cloud Pricing for Revenue Maximization , 2013, IEEE Transactions on Cloud Computing.

[12]  Liang Zheng,et al.  On the Viability of a Cloud Virtual Service Provider , 2016, SIGMETRICS.

[13]  Narendra Karmarkar,et al.  A new polynomial-time algorithm for linear programming , 1984, Comb..

[14]  Jimmy J. Lin,et al.  Large-scale machine learning at twitter , 2012, SIGMOD Conference.

[15]  Sergey Ioffe,et al.  Inception-v4, Inception-ResNet and the Impact of Residual Connections on Learning , 2016, AAAI.

[16]  Danny H. K. Tsang,et al.  M-Convex VM Consolidation: Towards a Better VM Workload Consolidation , 2016, IEEE Transactions on Cloud Computing.

[17]  Evgenia Smirni,et al.  CEDULE: A Scheduling Framework for Burstable Performance in Cloud Computing , 2018, 2018 IEEE International Conference on Autonomic Computing (ICAC).

[18]  Jun Li,et al.  ArA: Adaptive resource allocation for cloud computing environments under bursty workloads , 2011, 30th IEEE International Performance Computing and Communications Conference.

[19]  Baochun Li,et al.  Dynamic Cloud Resource Reservation via Cloud Brokerage , 2013, 2013 IEEE 33rd International Conference on Distributed Computing Systems.

[20]  Bhuvan Urgaonkar,et al.  BurScale: Using Burstable Instances for Cost-Effective Autoscaling in the Public Cloud , 2019, SoCC.

[21]  Evgenia Smirni,et al.  How to Supercharge the Amazon T2: Observations and Suggestions , 2017, 2017 IEEE 10th International Conference on Cloud Computing (CLOUD).

[22]  Sriram Ramabhadran,et al.  Cloud control with distributed rate limiting , 2007, SIGCOMM '07.

[23]  David Wentzlaff,et al.  Burstable Instances for Clouds: Performance Modeling, Equilibrium Analysis, and Revenue Maximization , 2019, IEEE INFOCOM 2019 - IEEE Conference on Computer Communications.

[24]  Haifeng Chen,et al.  Proactive Workload Management in Hybrid Cloud Computing , 2014, IEEE Transactions on Network and Service Management.

[25]  Costas Courcoubetis,et al.  Managing and pricing service level agreements for differentiated services , 1999, 1999 Seventh International Workshop on Quality of Service. IWQoS'99. (Cat. No.98EX354).

[26]  Sangtae Ha,et al.  Improving user QoE for residential broadband: Adaptive traffic management at the network edge , 2015, 2015 IEEE 23rd International Symposium on Quality of Service (IWQoS).

[27]  Theodore P. Hill,et al.  Finite-state Markov Chains , 2010 .

[28]  Randy H. Katz,et al.  Heterogeneity and dynamicity of clouds at scale: Google trace analysis , 2012, SoCC '12.

[29]  Andrew M. Odlyzko,et al.  Paris metro pricing for the internet , 1999, EC '99.

[30]  Giuliano Casale,et al.  Performance Prediction for Burstable Cloud Resources , 2016, VALUETOOLS.

[31]  George Kesidis,et al.  Using Burstable Instances in the Public Cloud: Why, When and How? , 2017, SIGMETRICS.

[32]  Evgenia Smirni,et al.  It's not a Sprint, it's a Marathon: Stretching Multi-resource Burstable Performance in Public Clouds (industry track) , 2019, Middleware Industry.

[33]  Zhi-Quan Luo,et al.  Semidefinite Relaxation of Quadratic Optimization Problems , 2010, IEEE Signal Processing Magazine.

[34]  Hojin Park,et al.  More IOPS for Less: Exploiting Burstable Storage in Public Clouds , 2020, HotCloud.

[35]  George Kesidis,et al.  Exploiting Spot and Burstable Instances for Improving the Cost-efficacy of In-Memory Caches on the Public Cloud , 2017, EuroSys.

[36]  Mohsen Guizani,et al.  Toward energy-efficient cloud computing: Prediction, consolidation, and overcommitment , 2015, IEEE Network.

[37]  Baochun Li,et al.  Reducing electricity demand charge for data centers with partial execution , 2013, e-Energy.