摘要

Ensuring data correctness over global clouds using classical pessimistic concurrency control approaches degrades the overall performance. To overcome such problem, a new breed of modern scalable approaches that adopt optimistic global concurrency control and/or non-serializable consistency is proposed. Such approaches cannot prevent conflicts as they allow replicas to diverge. Hence, they trade data correctness for performance. To overcome the limitations of the existing approaches, we propose an alternative novel approach (known as NASEEB) that uses the notion of a %26quot;Datacenter Escrow%26quot; to ensure data correctness over the cloud datacenters while maintaining good performance. NASEEB distributes the capacities of the data objects%26apos; attributes among the datacenters such that every datacenter has its own non-overlapping quota (i.e., the datacenter escrow). Each datacenter consumes its escrow locally without the need for any global checks. NASEEB allows datacenters to borrow quota from each other to meet incoming high demand. Experimental results show that NASEEB successfully maintains the data correctness as in the classical approaches and provides a good performance as in the modern approaches as long as the quota borrowing rate is kept low.

  • 出版日期2014-12