Home > Storage > PowerScale (Isilon) > Product Documentation > Management and Migration > Storage Tiering with Dell PowerScale SmartPools > Minimizing the performance impact of tiered data
Despite the performance enhancements that SmartPools can provide, any tiering approach will have some cases where performance can suffer. These degradations are the result of data location, tiering data movement activity, and so on.
Previously we discussed data location and performance isolation as performance enhancement functions of SmartPools. As expected, if data is on slower drive media, it will typically be slower to access.
Another architectural attribute of scale-out NAS is that data might reside on one node but the network access point for clients accessing that data may be on, or load balanced to, another class of node. This access node might have different front-end IO capabilities and cache, but the class of storage on which the data resides primarily govern the performance characteristics.
While SAS is usually faster than SATA, spindle counts can have a significant impact as well. The most important performance impacts in a SmartPools environment are the characteristics of the nodes in the node pool the application connects to, rather than what media type the data physically resides on. For example, assume that no bottleneck exists on the network side. A streaming application will experience little difference in performance between data on a SATA node with less CPU power, and data on a SAS node with more CPU power. This scenario is true if the application’s connection into the cluster is through a node pool where the nodes have a great deal of cache. A similar scenario exists for applications with random access patterns. If the node pool they connect into has sufficient CPUs, data location by spinning media type does not usually make a significant performance difference.
Another important performance consideration in a tiered storage environment is the effect of data movement itself on overall system resources. The action of moving data from one node pool to another uses system resources. SmartPools mitigates the impact of data movement in multiple ways. Architecturally, through the highly efficient Job Engine, and through impact policies. Impact policies are configurable by the end user to control the amount of system resources allocated to data movement and when data movement takes place. The Job Engine impact policies are: