Unified storage – or Multiprotocol storage – was a trend set by NetApp when it launched its FAS900-series storage subsystems in late 2002. Said to be fed up with watching potential business going to SAN vendors, the company introduced block-level access to its file-level products. Now all of its products can support both block-level and file-level storage.
Enterprises cite two main reasons for deploying unified storage, the first of which is server virtualisation. As dynamic data centres become a reality, they contain virtual machines (VMs) that run a range of operating and file systems that all address the same storage. Those VMs access storage in a variety of ways: at the file level using NFS or CIFS, at the block level using Fibre Channel (FC) or iSCSI, or even both ways. A unified storage system helps simplify deployment and removes barriers to the movement of VMs across the data storage infrastructure.
The second main reason to deploy unified storage is ease of management. Installing a single vendor's technology means there's only one set of objects to manage, which lowers operational expenditure, including maintenance and training. It also lowers risk because when drive capacity requirements are planned, there's no need to guess which storage pools will need upgrading.
As a result of such benefits, unified storage is gaining ground. Almost every storage vendor now sells unified storage products, and most enterprises either have or are developing plans to consolidate at least some of their network-attached storage (NAS) and storage-area network (SAN) resources. The most suitable area for consolidation is the second tier of storage in a hierarchical architecture, where applications aren't performance-critical and don't require dedicated resources. This also helps reduce the load on the SAN, which, as a performance-oriented technology, is better suited to serving applications such as databases than acting as end-user file storage.
Despite the benefits of unified storage, challenges remain for IT and storage managers wishing to implement it. One of the biggest hurdles is the problem of internal organisational politics: In most large enterprises, the network group is in charge of NAS while the storage group controls the SAN. Unifying the two raises significant managerial challenges for chief information officers (CIOs) and slows the process of unifying the storage infrastructure.
For analyst Tony Lock at Freeform Dynamics, there are other challenges. "All vendors sense the opportunities and have products," Lock said. "Whether that has translated into education of the market is debatable, especially among SMEs, where organisations might not have space or resources to run two types of storage, and usually have IT guys who do everything but not storage specialists."
Despite such roadblocks, for tier-2 applications the lines between SAN and NAS are likely to continue to blur, especially for SMEs. However, in large organisations, the distinctions between SAN and NAS will likely remain, because very demanding tier-1 applications such as transactional databases require the deployment of high-performance block-level storage.
Unified storage gets LOROL working
London Overground Rail Operations Ltd. (LOROL) runs train services across the UK capital and has a serious set of responsibilities for safety, punctuality and customer service. When formed in 2007 it needed a fast start-up, with only seven weeks to get systems up and running.
Unified storage wasn't initially part of the infrastructure picture, but it wasn't long before it became so, as LOROL's head of IT, Gareth Murphy, explained. "Before I talked to our supplier EACS and then NetApp, I was convinced we'd need a Fibre Channel-connected storage solution. Further demonstrations of iSCSI proved me wrong. The main driver for looking at such storage was a requirement for a new value-for-money product that needed minimal administration."
Murphy said the short period he had available to implement a system meant he didn't have much time to shop around. As a result of previous experience, he considered a Hewlett-Packard Enterprise Virtual Array (HP EVA) system but, he said, "It was vastly more expensive. Rather, I was pretty impressed with NetApp, which was also a lot cheaper."
The system LOROL acquired was a NetApp FAS2050, with 13 TB available, which has since been joined by a second unit. Each unit includes two controllers, one for CIFS to access file-level data and one for iSCSI to access block-level data. Some 60% of the storage is configured for iSCSI-addressed block storage, which is used by databases and LOROL's email archiving system. LOROL runs Exchange, SQL Server, Oracle and Autonomy Zantaz for email archiving, "and a small IBM database no one's ever heard of." The CIFS-addressed storage is used for end-users' Windows shares.
For Murphy, the advantages of the NetApp system were speed of implementation, ease of management and value for money. "Having experience of other vendors' products previously, NetApp offered a more affordable and, more importantly, easily scalable solution," he added.
Murphy said he liked that adding a second NetApp NAS unit just worked, with zero downtime. This characterised the implementation, as Murphy had no challenges to report. "Setting up from scratch gave us an opportunity to experiment with and test the environment before it was in production," he said, "which, along with NetApp's manageability, made implementing the storage solution one of the simpler aspects of setting up our infrastructure."
Murphy summed up the unified storage system as "a fast, value for money, very scalable storage solution that's easy to use and manage."