Amazon EFS quotas

Number of file systems for each customer account.

Elastic throughput quota per Regional file system for all connected clients in an AWS Region.

Provisioned throughput quota per Regional file system for all connected clients in an AWS Region.

Quotas for general resources, such as the number of access points or connections for each file system.

Elastic and Provisioned throughput quotas per One Zone file system for all connected clients in an AWS Region.

Bursting throughput quotas per Regional or One Zone file system for all connected clients in an AWS Region.

Maximum combined read and write throughput is 1,500 mebibytes per second (MiBps) for file systems using Elastic throughput and mounted using version 2.0 or later of the Amazon EFS client (amazon-efs-utils version) or the Amazon EFS CSI Driver (aws-efs-csi-driver). Maximum throughput for all other file systems is 500 MiBps. For more information about performance, see Performance summary. NFS client throughput is calculated as the total number of bytes that are sent and received, with a minimum NFS request size of 4 KB (after applying a 1/3 metering rate for read requests).

Up to 65,536 active users for each client can have files open at the same time.

Up to 65,536 files open at the same time on the instance. Listing directory contents doesn't count as opening a file.

Each unique mount on the client can acquire up to a total of 65,536 locks per connection.

When connecting to Amazon EFS, NFS clients located on-premises or in another AWS Region can observe lower throughput than when connecting to EFS from the same AWS Region. This effect is because of increased network latency. Network latency of 1 ms or less is required to achieve maximum per-client throughput. Use the DataSync data migration service when migrating large datasets from on-premises NFS servers to EFS.

The NFS protocol supports a maximum of 16 group IDs (GIDs) per user and any additional GIDs are truncated from NFS client requests. For more information, see Access denied to allowed files on NFS file system.

Using Amazon EFS with Microsoft Windows isn't supported.

pNFS

Client delegation or callbacks of any type

Mandatory locking

All locks in Amazon EFS are advisory, which means that read and write operations don't check for conflicting locks before the operation is executed.

Deny share

NFS supports the concept of a share deny. A share deny is primarily used by Windows clients for users to deny others access to a particular file that has been opened. Amazon EFS doesn't support this, and returns the NFS error NFS4ERR_NOTSUPP for any OPEN commands specifying a share deny value other than OPEN4_SHARE_DENY_NONE. Linux NFS clients don't use anything other than OPEN4_SHARE_DENY_NONE.

Access control lists (ACLs)

Amazon EFS doesn't update the time_access attribute on file reads. Amazon EFS updates time_access in the following events:

Namespaces

Persistent reply cache

Kerberos based security

NFSv4.1 data retention

SetUID on directories

Unsupported file types when using the CREATE operation: Block devices (NF4BLK), character devices (NF4CHR), attribute directory (NF4ATTRDIR), and named attribute (NF4NAMEDATTR).

Unsupported attributes: FATTR4_ARCHIVE, FATTR4_FILES_AVAIL, FATTR4_FILES_FREE, FATTR4_FILES_TOTAL, FATTR4_FS_LOCATIONS, FATTR4_MIMETYPE, FATTR4_QUOTA_AVAIL_HARD, FATTR4_QUOTA_AVAIL_SOFT, FATTR4_QUOTA_USED, FATTR4_TIME_BACKUP, and FATTR4_ACL.

An attempt to set these attributes results in an NFS4ERR_ATTRNOTSUPP error that is sent back to the client.

For a list of AWS Regions where you can create Amazon EFS file systems, see the AWS General Reference.

Amazon EFS does not support the nconnect mount option.

You can mount an Amazon EFS file system from on-premises data center servers using AWS Direct Connect and VPN. For more information, see Tutorial: Mounting with on-premises clients.