The EFS backend is designed for high performance and can deliver in excess of 7,000 file operations per second per filesystem. However, this performance cannot be maintained for long periods of time, as each EFS filesystem that we create will be throttled according to the burst credits that it has. Due to the spiky nature of network filesystem patterns, the system environment was designed similarly to T-type burstable instances. A baseline performance pattern has been determined and clients are allowed to burst over the baseline with very high rates of file operations, but they use up burst credits when they do so. Once the performance is within the baseline, the filesystem earns burst credits that can be used when needed. The allotment of the burst credits is determined by the size of the filesystem.