Andrew White, senior consultant at GlassHouse Technologies, looks at the benefits and challenges of unified storage -- subsystems that combine block and file access in one device. This interview is also available below as an MP3 download.
By submitting your personal information, you agree that TechTarget and its partners may contact you regarding relevant content, products and special offers.
Q. What is unified storage?
A. Unified storage is a storage system that's defined as being able to provide storage access at block level and file level. Block level has been traditionally Fibre Channel to a SAN [storage-area network], and now iSCSI adds to that as a method to allow servers and users to access data. Unifying storage brings those two access methods together under one single platform.
Q. What are the benefits of unified storage?
A. Unified storage addresses a couple of issues. One is that it allows a single platform for both file and block access. Traditionally, systems were islands where you had block-level data and you had file-level data, which meant two sets of infrastructure and two skillsets to manage the data. This brought not only challenges, but complexity in the data centre.
It also requires additional data centre resources to accommodate both, whereas with unified storage you have one platform. You have one system; one footprint in the data centre; you can realise power, space and cooling advantages; and it also unifies your management platform. Instead of running two or three different storage systems, you can now essentially run one and it allows the management of those systems to be much easier -- such as being able to report on capacity and other performance indicators. It makes administration a lot easier.
Technologically, most of these NAS [network-attached storage] systems came about by NAS vendors adding Fibre Channel or iSCSI block-level targets that allow the unification of block and file access.
Unifying those two platforms allows IT to provide the business with a greater level of service. If you have business requirements that allow both block and file access, you can turn around deploying those resources to the business quickly. It gives IT departments a great deal of flexibility in terms of meeting service-level requirements and that applies to external clients. too.
Q. What are the challenges of unified storage?
A. When considering unified storage, there are a number of concerns that need to be addressed before implementing platforms. Capacity management throughout the system is important. When you start mixing your unstructured typical NAS data with your block data, you're going to have to accommodate both data sets. We're seeing an explosive growth trend in unstructured data, which is traditionally held on these file-level server devices. Tracking and monitoring that data is important so you don't hit a boundary in terms of the capacity of the systems. That's definitely one concern, as it is with any storage system. If you have database systems that need to grow, and you're not able to do that because you've taken up all your space with file level-sharing, it can become an issue.
Performance is another issue. Block-level data and file-level data is accessed in different ways in the subsystem. Block-level data usually has shorter access times and is made up of shorter bits of data. File-level data is usually more random, can't take advantage of locality of reference and usually takes longer to service that request. Being able to determine how those different data sets interact with each other is important. Certainly, you would want to guarantee a certain performance level of your block-level data, [which is] typically your most critical data, such as your email applications, your online databases, your transaction processing -- applications that typically require a greater level of service than a file share.
If you're looking at applications that you want to merge onto a unified storage platform, typically it may be that you're better served by moving some applications that require higher I/O to a standard block storage system where you'd be able to tune the system appropriately so you can give proper precedence to those higher I/O, higher service-level requirements.
That's not to imply that unified storage systems are not well performing. That's not what I'm saying. They perform very well, but you have to understand your data access and data types before you can start consolidating storage arrays into one large, unified storage system.