WWR2 : Performance concepts and configuration : Introduction to performance modeling in Storage Modeller : Some ways that configurations affect performance predictions
Some ways that configurations affect performance predictions
The following table summarizes some configuration factors that affect performance.
 
Adapters
The number and type of I/O Adapters determine the point at which the adapter ports become a bottleneck in the system. Note that for pure capacity modeling, entering Adapter information is optional. However, for performance modeling, Adapter information is required. Storage Modeller generates an error message if it attempts to assign a workload to a system that lacks host adapters. The usage assigned to the adapter ports is also important, as it determines which ports, if any, are used for host I/O and for replication activity. Bottlenecks in the system can appear for either or both types of usage. The following factors are system-specific:
▪For IBM Storage Virtualize systems, adapter ports can also be used for node-to-node communication and for I/O operations to internal and external storage arrays. These can also present potential bottlenecks for the system.
▪For DS8000 systems, the number and types of device adapters used for I/O to the internal arrays is predetermined from the system type, model, and number of frames, and does not have to be explicitly entered by the user.
Pools
The number and type of pools configured for the system, the number and size of the arrays in the pools, and the type of drives in the arrays are significant determinants of system performance. Obviously, faster drives provide better overall performance.
Placing multiple tiers of drives into the same pool will signal the use of the Easy Tier function for IBM Storage Virtualize and DS8000 systems, which improves the performance for the pool. Configuring compression and/or deduplication for pools has significant benefit for capacity modeling, as it increases the effective capacity for the pools, but in some cases can have a negative effect on performance.
All these factors from the system configuration are taken into account when predicting the performance of the modeled systems. A prediction encompasses two main components:
▪The expected utilization of the system resources.
▪The expected response times of the running system.
The computed utilization values indicate which resources in the system are expected to become the bottleneck, limiting the total amount of I/O the system can process. In some cases, the following changes to system configuration can mitigate bottlenecks:
▪If the CPU (core or processor utilization) is the bottleneck, adding more CPU cores or more I/O Groups can potentially eliminate that bottleneck.
▪If the host interface is the bottleneck, adding more or faster I/O Adapters can potentially eliminate the bottleneck.
▪If the drives are the bottleneck, adding more or faster drives can potentially eliminate the bottleneck.