Performance Considerations

Farsight regularly issues updates to improve performance edge cases and make resource use more predictable. When given sufficient RAM - usually much less than the size of the batch being processed - Processing will work out of the box without any configuration. However, sophisticated users may want to optimize their workloads to get the most out of their hardware and can benefit from adhering to guidelines.

Small items of uniform size

For good CPU utilization it should be easy to divide work among the cores. Small items tend to have a more uniform size and are easier to assign to different cores.

Throughput depends on storage latency

File types with random access will be affected by underlying storage latency, increasing the network bandwidth may not lead to performance improvements. File types with linear access patterns, which read the file completely from start to finish are not affected by storage latency and will be processed faster with more bandwidth.

Tip

Consider copying files from higher latency storage to local storage for more predictable performance.

Pattern: Zipped PSTs

To optimize for size, PSTs may be stored in a ZIP or other compressed container format. Although this will decrease the size of PSTs making them more uniform, wrapping a PST in a compressed container will negatively affect performance, sometimes dramatically.

Tip

Consider running Processing on bare PST mailboxes.