Support for atomic writes

The Storage Foundation 6.2 release supports atomic write operations on RHEL6 on Fusion-io devices. Atomic write capable devices ensure that all blocks in write I/O operation (which may span multiple sectors) either pass or fail. If a write fails in-between, the storage reverts back to old data.

Atomic write resolves a problem of indeterminate status of failed writes that often requires two-part write - one write to an update log buffer and the other write on actual data volumes. Enabling atomic write eliminates the writes on log buffer, which in turn results in a better performance.

Storage Foundation lets you configure the atomic write support when you create a Veritas Volume Manager (VxVM) volume on a device that has atomic write capability. The atomic write I/O size of an atomic write capable volume is 16KB.

While creating an atomic write capable volume, VxVM ensures that all underlying subdisks are aligned to the 16KB boundary. Atomic write capable volumes can span multiple atomic write enabled devices, but I/O crossing atomic write boundary is not supported.

Atomic write is supported on raw VxVM volumes as well as on VxFS configured on VxVM volumes.

For information about using the Storage Foundation atomic write I/O feature with MySQL, see the Storage Foundation and High Availability Solutions Solutions Guide.