DQMP: A Decentralized Protocol to Enforce Global Quotas in Cloud Environments
Platform-as-a-Service (PaaS) clouds free companies of building infrastructures dimensioned for peak service demand and allow them to only pay for the resources they actually use. Being a PaaS cloud customer, on the one hand, offers a company the opportunity to provide applications in a dynamically scalable way. On the other hand, this scalability may lead to financial loss due to costly use of vast amounts of resources caused by program errors, attacks, or careless use.
To limit the effects of involuntary resource usage, we present DQMP, a decentralized, fault-tolerant, and scalable quota-enforcement protocol. It allows customers to buy a fixed amount of resources (e.g., CPU cycles) that can be used flexibly within the cloud. DQMP utilizes the concept of diffusion to equally balance unused resource quotas over all processes running applications of the same customer. This enables the enforcement of upper bounds while being highly adaptive to all kinds of resource-demand changes. Our evaluation shows that our protocol outperforms a lease-based centralized implementation in a setting with 1,000 processes.
KeywordsCloud Computing Fault Tolerance Cloud Provider Average Response Time Resource Controller
Unable to display preview. Download preview PDF.
- 1.Windows Azure Platform, http://www.microsoft.com/windowsazure/
- 2.Google App Engine, http://code.google.com/appengine/
- 3.Creeger, M.: Cloud computing: An overview. ACM Queue 7 (2009)Google Scholar
- 4.Schopf, J.M.: Ten actions when Grid scheduling: the user as a Grid scheduler. In: Grid Resource Management: State of the Art and Future Trends, pp. 15–23. Kluwer Academic Publishers (2004)Google Scholar
- 9.Uchida, M., Ohnishi, K., Ichikawa, K.: Dynamic storage load balancing with analogy to thermal diffusion for P2P file sharing. In: Proc. of the 2006 Work on Interdisciplinary Systems Approach in Performance Evaluation and Design of Computer & Communications Systems (2006)Google Scholar
- 10.Tassiulas, L., Ephremides, A.: Stability properties of constrained queueing systems and scheduling policies for maximum throughput in multihop radio networks. In: Proc. of the 29th IEEE Conf. on Decision and Control, pp. 2130–2132 (1990)Google Scholar
- 11.Xiao, L., Boyd, S., Lall, S.: A scheme for robust distributed sensor fusion based on average consensus. In: Proc. of the 4th Intl. Symp. on Information Processing in Sensor Networks, pp. 63–70 (2005)Google Scholar
- 13.Raghavan, B., Vishwanath, K., Ramabhadran, S., Yocum, K., Snoeren, A.C.: Cloud control with distributed rate limiting. In: Proc. of the 2007 Conf. on Applications, Technologies, Architectures, and Protocols for Computer Communications, pp. 337–348 (2007)Google Scholar
- 14.Pollack, K.T., Long, D.D.E., Golding, R.A., Becker-Szendy, R.A., Reed, B.: Quota enforcement for high-performance distributed storage systems. In: Proc. of the 24th Conf. on Mass Storage Systems and Technologies, pp. 72–86 (2007)Google Scholar
- 16.Hupfeld, F., Kolbeck, B., Stender, J., Högqvist, M., Cortes, T., Marti, J., Malo, J.: FaTLease: scalable fault-tolerant lease negotiation with Paxos. In: Proc. of the 17th Intl. Symp. on High Performance Distributed Computing, pp. 1–10 (2008)Google Scholar
- 17.Burrows, M.: The Chubby lock service for loosely-coupled distributed systems. In: Proc. of the 7th Symp. on Operating Systems Design and Implementation, pp. 335–350 (2006)Google Scholar
- 18.Weissman, C.D., Bobrowski, S.: The design of the Force.com multitenant Internet application development platform. In: Proc. of the 35th SIGMOD Intl. Conf. on Management of Data, pp. 889–896 (2009)Google Scholar
- 19.Douglas, S., Harwood, A.: Diffusive load balancing of loosely-synchronous parallel programs over peer-to-peer networks. ArXiv Computer Science e-prints (2004)Google Scholar