VAST Data announces per-job storage controls for VAST AI OS 5.6
Administrators will be able to cap bandwidth and I/O operations per second for each job sharing storage, VAST Data says. The feature requires VAST NFS 4.8 or later.
VAST Data announced Job-Level Quality of Service for VAST AI OS 5.6 on November 17 at SC26 in Chicago. The company says the forthcoming feature will let administrators cap bandwidth and input/output operations per second for each job using shared storage.
VAST says its NFS client attaches configurable job identifiers to storage requests so the storage system can apply the job’s policy. In VAST’s example, Slurm passes the job ID and quality-of-service setting to each job through its environment. VAST uses those identifiers to select the storage policy.
The company says the storage system can apply separate limits to jobs on the same compute node and one policy to the combined traffic of a job spread across several nodes. VAST says administrators can change read and write limits while jobs run, without restarting them, modifying applications or remounting storage.
VAST’s Slurm demo description covers a synthetic training workload with three jobs across two tenants. The announcement provides no independent measurements of how well the limits isolate jobs on shared storage.
In version 4.5.8, VAST added job ID support for NFS operations. VAST says the new controls require VAST NFS 4.8 or later and are planned for general availability with VAST AI OS 5.6 in late January or early February 2027.
More news

AWS introduces cross-Region inference routing for SageMaker HyperPod

AWS adds managed user administration to HyperPod Slurm clusters

Nscale says its first NVIDIA Vera Rubin NVL72 systems are online
