Requirement-Based Query and Update Scheduling in Real-Time Data Warehouses

摘要

A typical real-time data warehouse continually receives readonly queries from users and write-only updates from a variety of external sources. Queries may conflict with updates due to the resource competition and high loads. Moreover, users expect short response time for queries and low staleness for the query results. This makes it challenging to satisfy the two requirements simultaneously. This paper proposes a requirement-based querying and updating scheduling algorithm (RQUS) which allows users to express their real needs for their queries by specifying the acceptable response time delay and the acceptable result staleness when queries are submitted. RQUS dynamically adjusts the work mode of the system according to the changing requirements of users in order to allocate system resource to queries or updates and then prioritizes the query or update queue according to the work mode. And a freshness monitor is adopted to monitor the execution state of updating tasks in order to maintain the global table incrementally. Experimental results show that RQUS algorithm performs better than the three traditional scheduling algorithms with the changing user requirements overall.