Fibre channel dukes it out with IP: they're battling over cost and complexity.
Disk array growth was sluggish in 2001, but the market for storage software continued to grow as companies realized how very expensive manual maintenance is. According to Gartner, worldwide storage software revenue reached $4.9 billion in 2001, a 3% increase over 2000. EMC continues to lead the market, followed by Veritas and IBM. But Gartner points out that by excluding $1.6 billion in array-based software, Veritas would take the lead and EMC would move to third place behind IBM. EMC's software sales almost exclusively support its own arrays.
Hitachi Data Systems was a major success story. Combining excellent storage products and active partnerships with HP and Sun, it posted 59.4% growth numbers and made the Top 10 chart for the first time. Gartner also reported that pre-merger Compaq retained its top position in shipping worldwide disk-based storage and SANs for the second consecutive year.
Cost and Complexity
Fibre Channel still rules the waves, though there's a surging interest in IP-based storage area networks. Fibre Channel costs have dropped to $500 a port from $1,000, but equipment, consulting, installation, and ongoing management costs are still considerable. While SANs are widely accepted in corporate data centers, cost and complexity still hinder their growth outside the Fortune 1,500.
StoneFly Networks makes storage provisioning appliances for economical IP storage area networks. Speaking of lower costs for today's SANs, StoneHy Network's vice president of marketing, Jeffrey Schnabel, said, "The cost and complexity are the Achilles heel of storage area networks today. You just can't get there from Fibre Channel."
Rick Walsworth, director of marketing at Maranti Networks, said that the cost and complexity issues will limit SAN deployment beyond the Fortune 1,500. "SANs are fairly well employed in large data centers, because the value of data stored on the SANs is usually higher than the cost of the SANs and even the cost of managing them."
Walsworth believes that the true mission of modern storage management is to bring intelligence to networks so they can prioritize mission-critical applications without increasing complexity and cost. Maranti develops intelligent distributed switches.
Management interoperability remains a concern. There has been significant progress in the hardware area, with user demands driving interoperability development. The software and management side has not made as much progress, however. The ideal situation would be storage networks where all storage devices reported their status and other indicators to software management applications using widely accepted standard parameters. SNIA is championing CIM standards, and most storage vendors have come out in support of such CIM projects as the Bluefin initiative. Shared APIs, however, provide the best present approach to managing multivendor SANs. Peter Tarrant, vice president of marketing at file-based storage developer Agile Storage commented, "API specs are absolutely crucial to enabling a much more open environment." SNIA and the Bluefin standards, he continued, are vital. "Everyone has their role to play."
HP approaches storage as a utility. Mark Lewis, HP's worldwide head of marketing and solutions for the HP Networked Storage Solutions Group said, "We fundamentally believe in the concept of storage utilities, it's the way storage will evolve."
He cites two necessary actions:
* Decouple network storage from the server: Servers and their storage have been inseparable, so when one is replaced the other must be, too, regardless of how much life is still in it. Decoupling servers and storage allows administrators to manage them using independent life cycles.
* Apply storage technologies such as virtualization as a resource: Lewis pictures distributing these technologies to network users as similar to dumping those little power generators in favor of plugging into the power grid.
IBM's Scott Drummond, program director for storage networking, said IBM is focusing on helping users leverage current resources. "Along with the rest of industry, we're pushing TCO [total cost of ownership] models. That's just good common sense."
Companies identify stovepipes of server/storage, which are badly underutilized--as low as 15%--and often not well managed. "In direct-attached storage, you generally have ill-trained people and inconsistency in the way you treat your data across the organization. That's why we prefer to have networked storage."
Organizations tend to agree. Appalled at the low usage figures, they decide to consolidate their stored data onto storage area networks. But when users are required to give up their local storage and control, a successful transition to networked storage requires a tremendous shift in attitudes. Users must feel confident that their storage is a constantly available and simple utility that manages, presents, and secures their data.
Information security is a large part of the user trust equation. According to James Hughes, a fellow at StorageTek, "One of the impediments to widespread centralization of data is information security. Protection from destruction, protection from being able to destroy data--this is sometimes more important than keeping someone else from reading it."
Human considerations also factor into this area. "There are issues where organizations will have their data and they don't trust the IT department to have their data," said Hughes.
SANs Move to Mid-Tier
With slowing growth at the upper tier of business, most storage vendors are looking towards the mid-tier market, where at most only 25% of companies have adopted storage area networks. Given today's SAN cost and complexity, that's not likely to change--SAN adopters tend to have deep pockets, good IT departments and volumes of critical business information to protect. The cost is not just the steep initial purchase price, but also all the ancillary components that must go along with the servers and arrays-fiber, switches, routers, HBAs, tape drives. This very complexity is why consultants take up to a week to plan and install even very basic bundled SANs. When customers are used to setting up big servers in under an hour, they do not respond well to this time schedule.
But consolidating storage can still represent a good return on investment, even for midtier companies. LeftHand Networks manufacturers storage devices that attach to existing IP networks, and can serve either file-based or block-based data. The company primarily works with mid-tier companies. Dave Dupont, vice president of marketing and business development at LeftHand, said, "A SAN in some cases makes sense, but for most of the market, a device you can buy now that makes sense for less, that sounds like a good solution. I am not sanguine about accelerating growth in the SAN arena. They'll continue to be around, they have a customer base, but the very cost and complexity issues will eventually slow down that market opportunity. The reason for that is there are better approaches from a customer point of view."
Dupont sketches three common block-level storage architectures:
* Direct-attached storage (DAS): DAS is still the most common storage architecture and provides the server-based block level storage that many databases require. It is not centralized, though there are a growing number of products on the market that include DAS devices in data center storage discovery and management.
* IP SANs: IP-based SANs are generally the same architecture as a Fibre Channel SAN but run over IP. Many vendors are working in this market, citing users' existing familiarity with IP.
* Network-unified storage: This is LeftHand's term, though it is not alone in developing appliance-or module-based approaches to storage over IP networks for mid-tier markets.
A common approach to mid-tier market opportunities is bundling or "SAN-in-a-box," which lowers cost and complexity. IBM, for example, has worked with other members of SNIA's Support Solutions Forum (SSF) to put together four tested configurations, offering them to customers as guaranteed heterogeneous storage networks with cooperative service-level agreements between forum members.
IP-based SANs are seeing a lot of development and interest. However, not all IP storage vendors think iSCSI will topple Fibre Channel from its lofty perch at the top-tier companies, especially with complex SANs. iSCSI still requires expensive components such as TCP offload components, ASIC chips, and gigabit Ethernet cabling, as well as storage arrays and tape devices. And in answer to the claim that iSCSI will be easier to manage because IT administrators already know it, data network administration is a different animal from storage network administration. Managing storage over multiple servers is a much greater challenge than managing data over networks-storage administrators must now build zones, learn new volume management skills and adopt storage utilization approaches.
StoneFly's Schnabel thinks of IP SANs as a disruptive technology, a fascinating innovation that's unlikely to be mature enough at introduction to suit the very market it's aimed at-data-intensive and data-critical enterprises, the majority of which already have Fibre Channel SAN. StoneFly believes in IP-based SANs, but uses storage concentrators to hook up with customers' existing Ethernet equipment to create a storage area network.
What is on the horizon in SAN development? Agile's Tarrant cites three key trends in storage networking:
* Simplifying deployment of new storage networking technologies: Such as those that promote ease, automate usage configuration and maintenance, and allow incremental use of storage.
* Match application needs to storage: Easy set of tools to allow enterprises to match storage with application needs. For example, a critical, high-performance application needs a sophisticated storage approach, while a set of online training videos only requires a simple disk backup.
* Improve ROT: Help users get more return out of their SAN investment by using storage management applications to free up management resources, and pursue SAN/NAS convergence so users can transparently store both block-based and file-based data in the same storage networks.
As budgets tightened over the last year, storage vendors aggressively pushed lower-cost storage into both entrenched and emerging markets. Most customers looked to leverage their investment by adding lower cost storage and management applications. Appliance and module approaches may compliment or replace classic SANs in the mid-tier market, and NAS/SAN convergence will drive a good deal of storage development. Enterprise SANs may also become more common for the highest tier of SAN adopters as virtualization and storage automated management mature.
The future may well lie in a widening gap between the giant enterprise SANs expected in 2005 and economical, module-based SANs for the mid-tier markets. Tom Isakovich--founder, president, and CEO of TrueSAN--sees the majority of SAN deployments in this market. He likens the evolution of storage to that of mainframes in the data-networking world, where storage is following the same path from expensive, hard-to-manage, and monolithic storage, to less-expensive, scalable, and modular approaches.
Ease of use will become vitally important as smaller IT staffs come to depend on integrated storage software management suites instead of the dozens of SAN management tools they must use today. This is true of all storage environments, with a great deal of development happening around integrated suites.
|Printer friendly Cite/link Email Feedback|
|Author:||Chudnow, Christine Taylor|
|Publication:||Computer Technology Review|
|Article Type:||Industry Overview|
|Date:||Jul 1, 2002|
|Previous Article:||UWB rocks wireless: it's not a new technology, but ultra-wideband's commercial potential is finally getting noticed. (Business of Technology).|
|Next Article:||Blade or brick, take your pick: both increase server power, not server numbers.|