Metering: How Amazon EFS reports file system and object sizes

In the following section, find how Amazon EFS reports file system sizes and sizes of objects within a file system.

Metering Amazon EFS file system objects

Objects that you can view in an Amazon EFS system include regular files, directories, symbolic links, and special files (FIFOs and sockets). Each of these objects is metered for 2 kibibytes (KiB) of metadata (for its inode) and one or more increments of 4 KiB of data. The following list explains the metered data size for different types of file system objects:

• Regular files – The metered data size of a regular file is the logical size of the file rounded to the next 4-KiB increment, except that it might be less for sparse files.

  A sparse file is a file to which data is not written to all positions of the file before its logical size is reached. For a sparse file, in some cases the actual storage used is less than the logical size rounded to the next 4-KiB increment. In these cases, Amazon EFS reports actual storage used as the metered data size.

• Directories – The metered data size of a directory is the actual storage used for the directory entries and the data structure that holds them, rounded to the next 4-KiB increment. The metered data size doesn't include the actual storage used by the file data.

• Symbolic links and special files – The metered data size for these objects is always 4 KiB.

When Amazon EFS reports the space used for an object, through the NFSv4.1 space_used attribute, it includes the object's current metered data size but not its metadata size. You can use two utilities for measuring the disk usage of a file, the du and stat utilities. Following is an example of how to use the du utility on an empty file, with the -k option to return the output in kilobytes.

$ du -k file
4      file

Following is an example of how to use the stat utility on an empty file to return the file's disk usage.

$ /usr/bin/stat --format="%b*%B" file | bc
4096

To measure the size of a directory, use the stat utility. Find the Blocks value, and then multiply that value by the block size. Following is an example of how to use the stat utility on an empty directory:

$ /usr/bin/stat --format="%b*%B" . | bc 
4096

Metering an Amazon EFS file system

The metered size of an Amazon EFS file system includes the sum of the sizes of all current objects in the Standard and IA storage classes. The size of each object is calculated from a representative sampling that represents the size of the object during the metered hour. An example is the hour from 8 AM to 9 AM.

For example, an empty file contributes 6 KiB (2 KiB metadata + 4 KiB data) to the metered size of its file system. Upon creation, a file system has a single empty root directory and therefore has a metered size of 6 KiB.

The metered sizes of a particular file system define the usage for which the owner account is billed for that file system for that hour.

Note

The computed metered size doesn't represent a consistent snapshot of the file system at any particular time during that hour. Instead, it represents the sizes of the objects that existed in the file system at varying times within each hour, or possibly the hour before it. These sizes are summed to determine the file system's metered size for the hour. The metered size of a file system is thus eventually consistent with the metered sizes of the objects stored when there are no writes to the file system.

You can see this metered size for an Amazon EFS file system in the following ways:

• Call the DescribeFileSystems operation using one the SDKs, HTTP, or the AWS CLI.

• View the File Systems table, for each file system listed in the AWS Management Console.

• Run the df command in Linux at the terminal prompt of an EC2 instance.

  Use the df command and not the du command. Don't use the du command on the root of the file system for storage metering purposes. The results don't provide full data.

Note

The metered size of the Standard storage class is also used to determine your I/O throughput baseline and burst rates. For more information, see Throughput scaling with bursting mode.

Metering for infrequent access

Infrequent Access (IA) storage is metered in 4 KiB increments. IA file metadata (2 KiB per file) is always stored and metered in the Standard storage class. Data access for IA storage is metered in 1 MiB increments.

Metering provisioned throughput

Customers only pay for the amount of time that Provisioned Throughput is enabled. Provisioned Throughput is metered once every hour. For metering when Provision Throughput is set for less than one hour, Amazon EFS calculates the time-average using millisecond precision.

Amazon EFS file systems meter the throughput for read requests at one-third the rate of the other file system I/O operations. For example, if you are driving 30 MB/s of both read throughput and write throughput, the read portion counts as 10 MB/s of effective throughput, the write portion counts as 30 MB/s, and the combined metered throughput is 40 MB/s. This combined throughput adjusted for consumption rates is reflected in the MeteredIOBytes CloudWatch metric.