Reduced contention between queries and transactions

Fragmentation can reduce contention for data in tables that multiple queries and OLTP applications use. Fragmentation often reduces contention when many simultaneous queries against a table perform index scans to return a few rows.

For tables subjected to this type of load, fragment both the index keys and data rows with a distribution scheme that allows each query to eliminate unneeded fragments from its scan. Use an expression-based distribution scheme. For more information, see Distribution schemes that eliminate fragments.

To fragment a table for reduced contention, start by investigating which queries access which parts of the table. Next, fragment your data so that some of the queries are routed to one fragment while others access a different fragment. The database server performs this routing when it evaluates the fragmentation rule for the table. Finally, store the fragments on separate disks.

Your success in reducing contention depends on how much you know about the distribution of data in the table and the scheduling of queries against the table. For example, if the distribution of queries against the table is set up so that all rows are accessed at roughly the same rate, try to distribute rows evenly across the fragments. However, if certain values are accessed at a higher rate than others, you can compensate for this difference by distributing the rows over the fragments to balance the access rate. For more information, see Designing an expression-based distribution scheme.

Copyright© 2018 HCL Technologies Limited