Best practices: Proper sizing of cache disks

For best performance, the total disk cache size must be large enough to cover the size of your active working set. For read-heavy and mixed read/write workloads, this ensures that you can achieve a high percentage of cache hits on reads, which is desirable. You can monitor this via the CacheHitPercent metric for your S3 File Gateway.

For write-heavy workloads (e.g. for backup and archival), the S3 File Gateway buffers incoming writes on the disk cache prior to copying this data asynchronously to Amazon S3. You should ensure that you have sufficient cache capacity to buffer written data. The CachePercentDirty metric provides an indication of the percentage of the disk cache that has not yet been persisted to Amazon.

Low values of CachePercentDirty are desirable. Values that are consistently close to 100% indicate that the S3 File Gateway is unable to keep up with the rate of incoming write traffic. You can avoid this by either increasing the provisioned disk cache capacity, or increasing the dedicated network bandwidth available from the S3 File Gateway to Amazon S3, or both.