Broadly speaking, there are two ways to achieve unified storage. You can buy storage arrays that are built as dedicated multiprotocol access arrays with a NAS server and SAN controller combined under a single management interface. Or you can provide file-level access to a SAN disk array by adding a so-called NAS head or gateway.
Dedicated unified storage systems are popular at SMBs and remote offices where there is little flexibility in how servers are attached to storage. The unified storage head/gateway approach is better suited to enterprises, whose storage managers want more flexibility from their existing storage investments. However, this approach may result in storage administrators having to manage two systems: the SAN array and the NAS head.
Benefits of unified storage
Unified storage can be simple and inexpensive to manage, which is why such NAS-plus-block-level devices are popular for SME and remote office deployments. Using one device family across a business means there is only one platform to deal with, thereby making management and provisioning easier.
This flexibility also gives users time to develop the skills necessary to manage the various protocols. NAS systems are simplest to deploy and manage. Next comes NAS-plus-iSCSI, which adds an iSCSI driver that executes SCSI target-mode commands against a particular volume (or LUN) dedicated to a specific application. Those unified storage systems that add Fibre Channel capabilities are natively more complex because there are more high-end enterprise capabilities to contend with compared to an iSCSI environment, such as LUN masking or LUN affinity.
Unified storage can also bring cost savings, as the same disk array can be used for both file- and block-based access. This translates into a potential green storage benefit, since with NAS and SAN in one box, a company can cut down on floor space, power and cooling.
Challenges of unified storage
There are two key challenges with unified storage:
- the granularity of control achievable to deal with file and block access; and
- I/O issues that appear as devices try to handle the two types of traffic.
Although unified storage products allow two types of access methods, most of them have only a single cache to deal with both, with no ability to control the priority or caching amounts given to one method over the other. In addition, characteristics of data access, such as locality of reference and sequentiality, may differ greatly. File-based data often comprises longer streams of data and does not have the locality of reference found in systems dedicated to block-based access for heavy I/O transaction processing. Dedicated NAS and SAN products are tuned to dealing with these access patterns but most unified storage products are not. It's a good idea to check their ability to discern file-based access, apply dedicated resources and utilise optimised caching algorithms.
This lack of granularity of control over the I/O profile of the two types of access can add complexity. So while unified storage systems offer the promise of simplicity by incorporating two or more forms of access in one product, they may also bring management headaches.
A final challenge is the question of territory between IT department functions. SANs are usually managed by storage admins while NAS is often dealt with by the network or sys admin. Because unified storage straddles SAN and NAS, the issue of which group will oversee the unified storage system should be decided upfront.
This was first published in December 2008