Lately the demand for real-time data services has increased in applications where it is desirable to process user requests within their deadlines using fresh data. The real-time data services are usually provided by a real-time database (RTDB). Here, since the workload of the RTDBs cannot be precisely predicted, RTDBs can become overloaded. As a result, deadline misses and freshness violations may occur. To address this problem we propose a QoS-sensitive approach to guarantee a set of requirements on the behavior of RTDBs. Our approach is based on imprecise computation, applied on both data and transactions. We propose two algorithms to dynamically balance the workload and the quality of the data and transactions. Performance evaluations show that our algorithms give a robust and controlled behavior of RTDBs, in terms of transaction and data quality, even for transient overloads and with inaccurate run-time estimates of the transactions.
展开▼
