Software Defined Storage

Software-defined storage is latest buzz in storage industry. It gives rise to questions such as how software-defined storage is different from storage virtualization? Is this just old wine in new bottle?

Let us evaluate.

Storage started evolving since it came out of traditional computer systems and started getting deployed as external device. JBOD was simplistic external storage in early days. However, it came with couple of disadvantages: disk size was fixed and there was no protection against disk failures. This disk failure protection was then developed inside computer system as software in terms of RAID. However, since the computer system itself was vulnerable to failures, the disks in JBOD were still vulnerable to computer system failures resulting in data corruption. Next solution pushed the RAID software to HBA adaptors and then directly inside JBODs which were made more robust, failure tolerant and resistant. Thereafter all advanced features such as replication, snapshot, and deduplication started getting into JBOD converting them into sophisticated high end hardware storage arrays.

Though hardware storage arrays these days have many advance features, these features are applicable only at disk or logical unit (LU) level. Hardware arrays are not aware of contents within. This unawareness about content sometimes gives rise to inefficiency. E.g. many a times LUs of larger size are created for an application which doesn’t use that much of LU at least in beginning. Typically space required by application grows gradually. Till the time application uses full space of LU it is underutilized. Another example of inefficiency is storage arrays if capable of deduplication, may not be able to apply the dedup algorithm efficiently because contents of storage are not known.

In recent days this provisioning solved the space efficiency problem but no other problems such as object specific replication, deduplication, and snapshotting. In recent past cloud computing gave rise to need of multi-tenant storage. Multi-tenant storage with space efficiency still remains a challenge with traditional storage arrays.

Software-defined storage has evolved to address these problems by breaking LU as operational boundary. With software-defined storage, storage arrays will be able to receive content information and policies and apply them as appropriate.

Let us take an example of hypervisor application. Virtual machines are objects of hypervisor ecosystem and undergo various transitions during their life time based on application they are running. Apparently, data associated with VMs also needs to take part in this transition such as replication, snapshotting, cloning, and migration. If VM data is sitting on software-defined storage then these all requirement can be set as policies of data specific to VM.

Software-defined storage seems to be really addressing the issues that were unaddressed till date; hence it is not just old wine in new bottle.

Write to us at smm@calsoftinc.com

Contributed by: Tejas Sumant | Calsoft Inc.