The BeaST Grid as a Share-nothing Highly Available Storage

Mikhail E. Zakharov just announced a preliminary version of the BeaST Grid, a share-nothing highly available storage.

The BeaST Grid is the FreeBSD based reliable storage system concept. It utilizes two commodity servers into a pair of redundant active-active/asymmetric storage controllers which use iSCSI protocol (Fibre Channel in the future) to provide clients with simultaneous access to volumes on the storage backend which is also represented by three or more commodity servers. Using iSCSI (Fibre Channel in the future) both controllers have access to all storage nodes on the back-end. Depending on particular configuration it allows the BeaST Grid to create wide range of GEOM based software or hardware RAID array types along with ZFS storage pools.

One can create a bunch of independent machines, configure them together and act as a highly-available block storage. As a share-nothing architecture, there is no need to migrate any of the block storage or IP addresses between the machines. Most of the cloud services do not allow a single disk attached to two hosts simultaneously. Having two copies of data, attached to two hosts, is a simpler design, and it might be even cheaper in some cases.

For a more conventional setup where the disks are shared, refer to other BeaST implementations.

Leave a comment