My Amazon Elastic File System (Amazon EFS) file system had great performance immediately after I created it. Now, the performance is much lower. I want to know if this is related to burst credits.
File-based workloads typically spike and drive high levels of throughput for short periods, but lower levels of throughput for longer periods. Amazon EFS is designed to burst to high throughput levels for periods of time.
There are three throughput modes for your file system: Bursting Throughput, Provisioned Throughput, and Elastic Throughput. With Bursting Throughput, throughput on Amazon EFS scales as your file system grows. With Provisioned Throughput, you can instantly provision the throughput of your file system independent of the amount of data stored. With Elastic Throughput, you can scale your throughput up or down based on your workload.
All file systems have a permitted throughput that's the maximum throughput you can drive the file system at. The permitted throughput is either your baseline throughput or your burst throughput.
Both the baseline throughput and the burst throughput are based on the size of your file system. A metering process runs about every hour to determine the size of your file system. Based on the determined size, the process sets the baseline and burst throughput rates. When Amazon EFS Lifecycle management is turned on, only the size of the data in Standard storage class is considered.
Amazon EFS uses a credit system that determines when file systems can burst. If the credit balance of your file system drops to zero, then your permitted throughput rate drops to your baseline throughput. When you drive at baseline throughput, you use credits at the same rate that you earn them.
When you have a positive burst credit balance, you can drive your file system at its burst throughput. However, when you drive at burst throughput, you use credits faster than you can earn them. If this pace continues, then you use all the credits and your burst credit balance drops to zero.
New file systems start with 2.1 TiB of burst credits. This burst credit is why you can drive the new file system at burst throughput, even though no data or metadata is stored in the system. When the burst credits are used, the performance of the new system is lowered.
With Provisioned Throughput, you can instantly provision the throughput of your file system and get the high levels of throughput that your applications require.
Additional charges are associated with Provisioned Throughput mode. If you use Provisioned Throughput mode, then you're billed for the storage that you use and throughput that you provision independently. For more information, see Amazon EFS pricing.
To change between Provisioned Throughput mode and Bursting Throughput mode, you must wait more than 24 hours after the last throughput mode change. You can increase the Provisioned Throughput values as many times as you want, but you can decrease the Provisioned Throughput value only once every 24 hours.
If the metered size of your file system provides a higher baseline rate than the throughput you provisioned, then your file system defaults to the Bursting Throughput mode.
Depending on the workload, Elastic Throughput can automatically scale up or down for filesystem workloads that spike and are unpredictable. Use Elastic Throughput mode when you can't predict the workload of the file system. You can also use Elastic Throughput mode for applications that drive throughput at 5% or less of the average peak throughput. Because the throughput automatically scales, you don't need to provision the throughput capacity. When you use Elastic Throughput mode, you pay for the amount of metadata and read or write data and don't earn or spend burst credits.