SQL response time in parallel processing environments depends on several factors that do not arise in conventional systems. The first is intra-SQL parallelism: dividing the processing of a single SQL query into multiple subtasks that can be processed concurrently within a DBMS. As discussed in this paper, Oracle 10g, DB2 UDB ESE and Teradata all utilize different mechanisms to regulate intra-SQL parallelism, leading to special workload characterization problems in each case. SQL response time is also sensitive to usage patterns for rows, columns and individual data elements within a database. Data usage patterns play a critical role in determining the success or failure of certain performance optimization strategies such as constructing materialized views or creating additional indices. Workload characterization issues associated with SQL data usage patterns are also examined in this paper, with emphasis on data sources, analysis procedures, and applications.
展开▼