Regardless of whether you use the Pre-Migration Tool, the workload-migration process that occurs as part of upgrading to SLES 11 specifies system-wide limit and reserve values for the Always planned environment, as shown in the following table.
|CPU Limit||Existing value in "PsfGlobalOpenvValues" of base state (Always PlannedEnv) otherwise 100|
|Reserved AWTs||Existing value in "PsfGlobalOpenvValues" of base state (Always PlannedEnv) otherwise 0|
|Max AWTs for new work||Existing value in "PsfGlobalOpenvValues" of base state (Always PlannedEnv) otherwise 80|
- In SLES 11, operating system utilities and work running on nodes external to those housing the Teradata Database are not subject to the CPU limit. However, non-database demands as well as resources that may be required for tuning or growth purposes should be taken into account when establishing CPU limit. By adjusting the limit downward from 100%, you can provide adequate resources for all work while reserving resources for any necessary refinements.
- All database activities that were exempt from the CPU limit in SLES 10, including rollbacks when running at the default DBS control settings and work running in the System performance group, are subject to the limit in SLES 11 by default.
- Unlike in SLES 10, tactical workloads in SLES 11 are subject to the CPU limit. However, in SLES 11, such workloads are allocated the highest priority by default. Thus, there is no need to limit CPU at the system level to insulate tactical work from competition for CPU resources.
- To protect against a perceptible increase in average elapsed time for completing highly critical work or work associated with short service level agreements, CPU limit be should be set no lower than 63%.
- Only physical I/Os that actually involve reading or writing from disk are subject to I/O limit. The cost of each such I/O is determined after the I/O request arrives, based on hardware specifications for the device. This cost is then subtracted from the total allowable bandwidth for the device as mandated by the established I/O limit value. A given device’s available I/O bandwidth can be ascertained at any point in time based on the known maximum bandwidth of the device and the cost of the I/Os already performed. When I/O limit is reached within a given enforcement interval, I/O activity on the device ceases for the remainder of that interval.
For more information, see Teradata Viewpoint User Guide. Additional information pertinent to CPU and I/O limits specifically can be found in the following Orange Book: Workload Management Capacity on Demand, Teradata Database 14.10 and Linux SLES11, 541–0010245–A02.