Published release of the cluster file system Lustre 2.17, used in most of the largest Linux clusters containing tens of thousands of nodes. The key components of Luster are metadata processing and storage servers (MDS), management servers (MGS), object storage servers (OSS), object storage (OST, supports running on top of ext4 and ZFS) and clients. The project code is distributed under the GPLv2 license.
Main innovations:
- Enabled by default is a hybrid I/O performance optimization mechanism that automatically applies buffered or direct I/O mode depending on the nature of the load and the size of the data transferred.
- Added support for dynamic configuration of the Nodemap engine, which is responsible for mapping user identifiers (UID/GID) in “multi-tenancy” configurations, in which different groups of clients have access to the FS.
- In the dynamic configuration system LNet NID (LNet Network Identifier) implemented Correct processing of attempts to register more than 32 NIDs on the server (previously, a similar situation led to a failure, but now NIDs in excess of
/Reports, release notes, official announcements.