Storage Management

The MASS Dashboard provides a common place to manage storage resources, client access, and system health. Applications use the resulting storage connections to read and write data directly.

Storage Resources

For file and object storage, a volume group defines the capacity available to its volumes. A volume provides the data space an application uses, with its own access settings and data protection configuration.

A KV cache store provides a shared cache for a set of GPU servers. Its configuration identifies the participating storage nodes, cache page size, and servers allowed to connect. It uses a separate data path from file and object volumes.

Client Access

Administrators configure identities and access permissions in the Dashboard. On a client machine, mass-ctl handles authentication and connection setup: mounting a filesystem volume or connecting a GPU server to a KV cache store.

Connecting a GPU server makes the store available to its cache software. The inference engine's integration determines which KV cache pages to save and when to retrieve them.

Resource Lifecycle

Creating or deleting storage involves background work. The Dashboard shows the resource's status so the administrator can see when it is ready or whether an operation needs attention.

These management operations are separate from application I/O. Changing a storage resource or its access settings can still affect the clients using it; separating the paths does not make such changes invisible to applications.

Monitoring

Capacity, performance, and health views help administrators understand usage and investigate changes in behavior. KV cache statistics include cache hits and misses, page reads and writes, and activity by connected server.

Use these measurements together: a high cache hit ratio describes reuse, while read latency and application response time show how that reuse affects inference.