The tracking of I/O cost limits may be less precise for partitioned tables than for unpartitioned tables, when Adaptive Server is configured for parallel query processing.
When you query a partitioned table, all the labor in processing the query is divided among the partitions. For example, if you query a table with three partitions, the query’s work is divided among 3 worker processes. If the user has specified an I/O resource limit with an upper bound of 6000, the optimizer assigns a limit of 2000 to each worker process.
However, since no two threads are guaranteed to perform the exact same amount of work, the parallel processor cannot precisely distribute the work among worker processes. You may get an error message saying you have exceeded your I/O resource limit when, according to showplan or statistics io output, you actually have not. Conversely, one partition may exceed the limit slightly, without the limit taking effect.
See the System Administration Guide for more information about setting resource limits.