FAQ

Unified storage architecture: Different approaches, scalability and capacity-saving tools

A unified storage architecture provides file and block storage access in the same device with connections via Fibre Channel and/or iSCSI and CIFS/NFS depending on the product. Putting the two types of storage access together potentially cuts capital costs and eases management but you need to watch out for performance bottlenecks.

In this interview, SearchStorage.co.UK Bureau Chief Antony Adshead speaks with Ian Lock, service director for storage and backup with GlassHouse Technologies (UK), about the benefits and limitations of a unified storage architecture, how vendors’ unified storage products differ, and how scalable and flexible it is.

Read the transcript below or listen to the unified storage architecture podcast

SearchStorage.co.UK: What is a unified storage architecture, and what are its benefits and limitations?

Lock: A unified storage system is one which provides both block- and file-based data access from a single set of storage controllers and from a single pool of storage. Typically, the block-based component is served up to hosts using Fibre Channel or iSCSI, whereas file-based data access is provided by CIFS or NFS file serving protocols.

The benefits come from the consolidation of multiple different storage arrays into one platform with a single pool of disk drives and a single management interface, which requires just a single skill set from your storage admins.

The limitations, as always, when you consolidate formerly separate systems into one come potentially from bottlenecks as you put more of your storage eggs in one basket.

A sound design with a good understanding of the performance requirements of your users and hosts becomes even more important than ever.

SearchStorage.co.UK: What different approaches to unified storage are there?

Lock: The different approaches to unified storage largely depend on the background of the storage vendor. Some vendors come very much from a block storage background. That is, their heritage for the last 15 or so years has been in producing a combination of monolithic or modular Fibre Channel-based storage systems, which hosts connect to using Brocade, Cisco or QLogic Fibre Channel switches and host bus adapters.

On the other hand, some vendors approach things from a file sharing perspective, advocating data access for their storage systems using Ethernet networks and CIFS or NFS.

Both approaches have their benefits. Fibre Channel, block-based storage typically provides the best performance for transactional systems like databases and messaging systems, but their Fibre Channel networking requirements can make them more costly.

Ethernet-based file protocol systems typicallyhave lower overall costs and make more sense for consolidating unstructured file data from disparate Windows and Unix file servers but are often seen as less performant due to the overhead of the Ethernet protocol. [However,] the smaller, niche providers of clustered NAS have sought to architect around these limitations with multiple connected processing units.

Both sets of vendors realised pretty early on that they were missing out by not providing products which played both in block and file markets. So, block storage vendors started shipping NAS head units, which allowed file access to their back end block storage, and the file storage vendors started using iSCSI to allow block-based access to their storage over Ethernet. They later allowed direct Fibre Channel connection to their NAS-based units.

So, at this stage both sets of vendors were providing what I call prototype unified storage systems. Since then both sets of products have continued to evolve, in particular introducing support for Fibre Channel over Ethernet (FCoE).

Today, storage vendors talk in terms of true unified storage having a single management interface, a single set of data management policies for both file and block storage and a single common storage pool.

Read the transcript below or listed to the unified storage podcast.

SearchStorage.co.UK: How flexible is a unified storage architecture? How easy is it to scale such systems and to add or remove protocols?

Lock: Unified storage is as flexible and scalable as its underlying operating system and hardware architecture allows. The protocols used are defined by the operating system. In general, those are Fibre Channel, NFS, CIFS, iSCSI and maybe FCoE.

The latest generation of unified storage systems includes software features to help flexibility and scalability, such as thin provisioning, which allows us to drive up utilisation levels in the storage pool; deduplication and compression, which can increase the effective storage capacity of the storage array by shrinking data and files; and dynamic data migration, to move data to the right tier of storage based on its profile.

The latest generation of unified storage systems can support from tens to over 1,000 disk drives and use solid-state flash drives, SAS and SATA drive technology. This lets you scale from a few terabytes to petabytes of capacity.

Most vendors also ship gateway versions of their products, allowing their intelligent processing or head units running their unified storage operating systems to use different back-end storage from third-party vendors. This makes good sense as in many enterprise environments companies have built up an estate of arrays from many vendors over the years and cannot afford to simply discard them.

In summary, unification of block and file arrays, along with bringing storage and compute (or server) environments close together, is the next macro-trend for storage.


This was first published in April 2011

 

COMMENTS powered by Disqus  //  Commenting policy