tailieunhanh - Deploying Database Appliances in the Cloud

Once an agreement about how to handle the transaction has been reached, it must be executed to completion according to the plan— node failure and related problems cannot cause the transaction to abort. If a node fails, it can recover from a replica that had been executing the same plan in parallel, or alternatively, it can replay the history of planned activity for that node. Both parallel plan execution and replay of plan history require activity plans to be deterministic—otherwise replicas might diverge or history might be repeated incorrectly. To support this determinism guarantee whi le maximizing con- currency in transaction execution, Calvin uses a deterministic lock- ing protocol based on. | Deploying Database Appliances in the Cloud Ashraf Aboulnaga Kenneth Salem Ahmed A. Soror Umar Farooq Minhas Peter Kokosielis Sunil Kamatht University of Waterloo IBM Toronto Lab Abstract Cloud computing is an increasingly popular paradigm for accessing computing resources. A popular class of computing clouds is Infrastructure as a Service IaaS clouds exemplified by Amazon s Elastic Computing Cloud EC2 . In these clouds users are given access to virtual machines on which they can install and run arbitrary software including database systems. Users can also deploy database appliances on these clouds which are virtual machines with pre-installed pre-configured database systems. Deploying database appliances on IaaS clouds and performance tuning and optimization in this environment introduce some interesting research challenges. In this paper we present some of these challenges and we outline the tools and techniques required to address them. We present an end-to-end solution to one tuning problem in this environment namely partitioning the CPU capacity of a physical machine among multiple database appliances running on this machine. We also outline possible future research directions in this area. 1 Introduction Cloud computing has emerged as a powerful and cost-effective paradigm for provisioning computing power to users. In the cloud computing paradigm users use an intranet or the Internet to access a shared computing cloud that consists of a large number thousands or tens of thousands of interconnected machines organized as one or more clusters. This provides significant benefits both to providers of computing power and to users of this computing power. For providers of computing power the push to cloud computing is driven by economies of scale. By operating massive clusters in specially designed and carefully located data centers providers can reduce administrative and operating costs such as the costs of power and cooling 15 16 . In addition the per-unit costs of