Limits and performance metrics of Apsara File Storage for HDFS-Apsara File Storage for HDFS(HDFS)-阿里云帮助中心

This topic describes the limits and performance metrics of Apsara File Storage for HDFS.

Performance metrics

Metric	Description
Data durability	99.999999999% (eleven 9s)
Average 4 KB read latency	Less than 1 ms
Average 4 KB write latency	Less than 1 ms
Maximum throughput	150 GB/s
Maximum read and write throughput density	150 MB/s/TB
Maximum IOPS	10,000 per TB

Limit	Limitations
File systems per account per region	3
Mount targets per file system	2
Concurrent compute node mounts per file system	Unlimited
Maximum directory depth	1000 layers
Maximum files or directories per file system	1 billion
Maximum capacity per file system	1 PiB
Permission groups per account per region	10
Permission rules per permission group	100
Concurrent active processes per mount target	Supports 10,000 active SDK processes

Symbolic links are not supported.
File concatenation (concat) is not supported.
Snapshot-related operations are not supported.
Operations related to delegation tokens and checksums, such as setWriteChecksum and setVerifyChecksum, are not supported.
ACL-related operations are not supported.
FileBlockLocations are not supported.
Path, file, and directory naming conventions
- Paths
  - Only Unicode characters and strings are supported.
  - Path elements cannot contain ASCII characters from 0 to 31.
- Filenames and directories:
  - The value cannot be an empty string, a single dot (.), or two consecutive dots (..).
  - Cannot contain colons (:), forward slashes (/), or tildes (~).

Snapshot-related commands, such as createSnapshot, deleteSnapshot, and renameSnapshot, are not supported.
ACL-related commands, such as setfacl and getfacl, are not supported.

For more information about open source Hadoop, see the Hadoop documentation.

The WebHDFS REST API is not supported.