Posted
Feb 1, 2003
 | By
David Braue

Truce in the storage battle

The storage industry has an interesting problem. Once limited by the fact that customers had to buy additional, expensive servers to add more storage space, the industry's growth accelerated with the popularity of network-attached storage (NAS) - which enabled scalability without the complexity of managing additional servers.

Yet as NAS devices proliferated, they presented similar problems: large volumes of disk required careful management that became difficult with every new filer that was added. Furthermore, increased utilisation of centralised storage imposed a heavy burden on networks already struggling to move all the data the company needed.

Resolving this problem was the goal of storage area networks (SANs) but their high cost, still-evolving standards, nagging interoperability issues and the need for new skills and specialised equipment - including Fibre Channel switches and fibre-attached storage arrays - have so far limited their adoption. The move from file-based storage to block-based storage simply hasn't been smooth or simple enough for most. SANs' potential has also been restricted by the general economic slowdown, making companies hard-pressed to justify the rip-and-replace approach necessary to move into the SAN world.

Momentum has tended to keep most companies maintaining the same storage strategy, with loads of direct attached storage (DAS) still going into the market despite recognition that NAS is easier to manage. Nonetheless, the price per megabyte has been pushed down to record lows; this has seen storage revenues fall sharply even though the number of gigabytes shipped is steadily increasing. According to research company IDC, total world storage industry revenues dropped 2 per cent, to US$4.7 billion, in the third quarter of last year alone.

That presents storage's giants - Hewlett-Packard, IBM, EMC, Dell and Hitachi round out IDC's top five - with a serious conundrum: how can they convince customers to buy more storage, while promoting newer architectures and products where there's room for profit-producing value add?

Their answer has been to step away from the religious wars pitting SAN against NAS, and instead to focus on new ways that customers can tailor their disk arrays to their individual needs.

Burying the hatchet

Making this step first required vendors to clean up the mess they'd made of interoperability; in most discussions, NAS and SAN philosophies had become so polarised that many customers had come to fear committing to one or the other. This put the squeeze on companies like Network Appliance, whose traditional strength has been in NAS. With customers recognising the value of a SAN, NetApp needed a way to play in both camps.

The company's solution? A hybrid 'unified storage' box that uses a native block-based architecture to support both file-based NAS environments and SAN storage. Storage managers can split the disk space between architectures as necessary, allowing file-and-print environments to store data in conventional NAS fashion while retaining the transactional performance benefits of SANs' block-based structure where necessary.

"Organisations have typically upgraded by forklift," says NetApp Australia-New Zealand Marketing Manager Harry Christian. "We're saying 'buy the filer today, and it will chop and change as the business demands'. It has simplified storage without having to go through the decision making process and being lumbered with the decision for years."

Whether NetApp's fence-sitting will pay off in the long run remains to be seen, but it's an indication of the growing trend to stop treating NAS and SAN as being worlds apart.

Making this happen has required considerable effort in translating between storage control protocols. For example, FICON (FIber CONnector) interfaces translate between mainframes' ESCON signalling and the Fibre Channel used in SANs. This allows mainframes to be hooked into conventional SANs.

Other standards, such as iSCSI (which sends disk arrays' SCSI commands over standard IP networks) and iFCP (Internet FCP, which translates between Fibre Channel commands and standard IP networks), allows customers to extend SANs over public or private IP networks anywhere in the world - without going to the expense of long-distance Fibre Channel runs. This approach has enabled the creation of trans-oceanic SANs with little or no performance hit.

Using standards such as IP and Gigabit Ethernet - as opposed to relying on expensive Fibre Channel - allows customers to support and extend SANs without being locked into a storage ivory tower. Rather than being forced to replace existing storage systems, existing systems can be linked into the SAN to obtain benefits such as easier back-up, no consumption of general network bandwidth and centralisation of enterprise storage.

Yet while the NAS and SAN worlds are being bridged at the protocol level, there is still a lot of work to be done before SANs are as easy to use as conventional filers. Much of this work comes from the need to ensure that management software is capable of addressing every element within a SAN - a goal that's been elusive thanks to years of proprietary and incompatible SAN hardware. Compatibility between vendors has been driven less by standards than by strategic alliances in which storage providers certify specific combinations of equipment.

That's not to say that the differences can't be surmounted: FalconStor, for one, filled the interconnectivity niche with its IPStor architecture, which translates data between SCSI, Fibre Channel and IP. By wrapping this multi-protocol support into a full suite of data management tools - including snapshot copying, backup tools, multi-pathing, mirroring, bare metal recovery, remote booting and more - IPStor allows management of all storage devices from a single console.

It can also integrate SANs with existing NAS environments. "We're trying to be the Switzerland of the software world," says FalconStor co-founder Dr Eric Chen.

The need to rely on third-party software for interoperability will soon be eliminated by the introduction, potentially this year, of CIM (Common Information Model), a standard defining a common interface to all kinds of storage equipment. Broad support for CIM, along with its derivative WBEM (Web-Based Enterprise Management), will allow management software to better interface with any type of storage equipment without requiring pre-certification of that equipment's compatibility.

New storage building blocks

While many vendors have worked on software to bridge NAS and SAN, others have focussed their efforts on packaging storage in new ways.

One popular approach has been a growing reliance on inexpensive ATA disks - the kind found in conventional desktop PCs. Because they're cheaper than high-performance SCSI drives, ATA drives (which cost just a few cents per megabyte) are never going to be used as primary storage devices; they're just too slow and lack the reliability of their more expensive cousins.

Where they are finding a home, however, is as a replacement for tape, whose high capacity and low cost has made it the universal backup standard. Products such as NetApp's NearStore R100, however, seek to supplant tape by aggregating massive numbers of cheap drives into a single, manageable box providing large amounts of 'near-line' storage.

The R100, for one, provides from 12 to 96 TB of storage. And while it's slow, it's fast enough to house one or more full copies of most companies' data - and those back-ups can be made without ever switching a tape. Better still, near-line storage remains available all the time. That characteristic has helped near-line storage devices fuel a renaissance in hierarchical storage management (HSM) - a mainframe-world technique in which less frequently-used data is moved from primary to secondary storage (previously, a tape drive).

Using near-line storage and an effective HSM application, it's easy to keep mission-critical data on faster disk arrays while moving archived information onto the slower ones. Targeting the almost viral growth of email within most companies - and the increasingly onerous legal obligations that accompany it - companies like Legato, Ixos, StorageTek and Veritas offer tools allowing customers to define policies that automatically age and move old emails from users' inboxes onto near-line storage. There, they remain accessible to users but don't clog up desktop PCs or consume unnecessary space during backups.

The argument for cheap disks is so compelling that even Quantum, a long-time advocate and producer of tape drives, has jumped into the fray. Its DX30 looks like an ATL P1000 tape drive to the network, but actually contains 3.8 TB of inexpensive disks that can store data at around 80 MB per sec - five times faster than DLT tape.

"This can be installed in about half an hour, utilising exactly the same backup processes and procedures that customers already have licences to support and people trained on," says Mike Sparkes, Quantum's Asia-Pacific Product Marketing Manager. "Using disk takes out a lot of the management costs."

Cheap disks aren't the only new thing in storage, however; some companies are building boxes that are smarter as well as bigger. EMC, for example, has built content awareness into its self-healing Centera storage boxes, which attach tags to specific types of data and use those tags to reference and authenticate that data within the archive.

EMC's Celerra CNS is also content-aware, working with EMC's OnCourse software to monitor the content stored in the box. This allows the storage system to store certain types of data in a certain way, or to proactively report on published data - for example, emailing updates of a particular document according to programmed business rules.

StorageTek has taken an entirely different tack on reducing management costs, borrowing a page from server vendors to build its next-generation BladeStore disk subsystems with a modular blade architecture. Instead of buying a fully populated filer, buying BladeStore chassis provides a housing for anywhere from 8 to 160 TB of online storage space that is expanded by simply installing additional blades, or interface cards. The chassis presents itself as an array of Fibre Channel drives.

There are many other options when it comes to revamping your storage strategy. If you're looking for performance rather than scalability, consider adding a solid-state drive like Platypus Technology's QuikData, which combines a SDRAM-based cache with uninterruptible power supply to speed server performance markedly.

If price is the biggest factor, consider the wealth of SME-focused storage boxes now on the market. Commodity storage favourite Iomega recently launched a range of SME NAS boxes offering from 160 to 480 GB of storage, while EMC's Celerra NS600 was designed to undercut NetApp in the NAS arena.

With a wealth of options now on the market, there's no need to get sucked into the NAS vs SAN debate. Today's storage is smarter and bigger than ever before; shop around and find out what the storage vendors will do for you. You may be surprised just how much it improves your data management.