Managing throughput capacity on FSx for Windows File Server file systems
Throughput capacity is one factor that determines the speed at which the file server hosting your FSx for Windows File Server file system can serve file data. Higher levels of throughput capacity also come with higher levels of I/O operations per second (IOPS) and a larger amount of cache memory on the file server. You can modify your file system's throughput capacity to help manage its performance at any time. For more information, see FSx for Windows File Server performance.
Topics
How throughput scaling works
When you modify your file system's throughput capacity, Amazon FSx switches out the file system's file server behind the scenes. For Multi-AZ file systems, switching to a new file server triggers an automatic failover and failback while Amazon FSx switches out the preferred and secondary file servers. Single-AZ file systems will be unavailable for a few minutes while the file server is switched during throughput capacity scaling. You are billed for the new amount of throughput capacity once it becomes available to your file system.
Note
During a maintenance operation on the back end, system modifications (such as a modification to your throughput capacity) may be delayed. Maintenance can cause these changes to queue up until they are next to be processed.
For Multi-AZ file systems, throughput capacity scaling results in an automatic failover and failback while Amazon FSx switches out the preferred and secondary file servers. During file server replacements, which happen during throughput capacity scaling as well as file system maintenance and unplanned service disruption, any ongoing traffic to the file system will be served by the remaining file server. When the replaced file server is back online, FSx for Windows will run a resynchronization job to ensure that data is synced back to the newly replaced file server.
FSx for Windows is designed to minimize the impact of this resynchronization activity on application and users. However, the resynchronization process involves synchronizing data in large blocks. This means that a large block of data can require synchronization even if only a small portion is updated. Consequently, the amount of resynchronization depends not only on the amount of data churn, but also the nature of the data churn on the file system. If your workload is write-heavy and IOPS-heavy, the data synchronization process may take longer and require additional performance resources.
Your file system will continue to be available during this time, but in order to reduce the duration of data synchronization, we recommend modifying throughput capacity during idle periods when there is minimal load on your file system. We also recommend ensuring that your file system has sufficient throughput capacity to run the synchronization job in addition to your workload, in order to reduce the duration of data synchronization. Lastly, we recommend testing the impact of failovers while your file system has a lighter load.
Knowing when to modify throughput capacity
Amazon FSx integrates with Amazon CloudWatch, enabling you to monitor your file system's ongoing throughput usage levels. The performance (throughput and IOPS) that you can drive through your file system depends on your specific workload’s characteristics, along with your file system’s throughput capacity, storage capacity, and storage type. You can use CloudWatch metrics to determine which of these dimensions to change to improve performance. For more information, see Monitoring with Amazon CloudWatch.
FSx for Windows File Server provides performance alerts based on values of CloudWatch metrics for your file system in the Monitoring & performance dashboard in the File system details page on the Amazon FSx console. This includes throughput capacity, and other file system metrics that can benefit from throughput capacity increases. For more information, see Performance warnings and recommendations.
Configure your file system with sufficient throughput capacity to meet not only the expected traffic of your workload, but also additional performance resources required to support the features you want to enable on your file system. For example, if you’re running data deduplication, the throughput capacity that you select must provide enough memory to run deduplication based on the storage that you have. If you’re using shadow copies, increase throughput capacity to a value that's at least three times the value that's expected to be driven by your workload to avoid Windows Server deleting your shadow copies. For more information, see Impact of throughput capacity on performance.