Storage Trends and Directions

As the Millennium approaches, most industy watchers agree that the cost to companies from Y2K bugs will be quickly dwarfed by the "discovery" within many corporations of the true cost of storage management. Centralization strategies with enhanced management automation are a prerequisite for preventing the storage explosion of the 1990s from becoming the IT disaster of the 2000s.

To Dave Duren, Manager of Information Facilities for lawn maintenance equipment manufacturer, Simplicity Manufacturing (Port Washington, Wis.), the key to planning a strategic storage architecture is centralization. In January 1998, Duren and his six-person IT staff deployed a Unisys NX4800 ClearPath mainframe and an EMC 5400 Symmetrix storage array. The solution allowed the consolidation of some NT server-based applications onto the "NT side" of the ClearPath mainframe and enabled the consolidation of storage.

Says Duren, the change from an older mainframe and storage devices to the newer units freed up substantial floor space in the corporate data center, setting the stage for server re-centralization. "It took us down to about 32 square feet of equipment space in the computer room, so we have been able to begin bringing servers back into the data center."

Duren observes that the selection of the EMC storage array was a key enabler of the strategy, "We bought the EMC platform with about 144GB of mirrored RAID disk. A little over 90GB are usable. Of that amount, about 72GB of storage is used for ClearPath mainframe-side applications and the balance is used for NT-side storage. The unit isn’t even half filled with drives. We could literally triple the amount of storage, given current drive sizes. There is a lot of room for growth and room for centralizing the storage associated with our distributed NT and Novell servers, if we decide to go that way."

Duren says that the decision to use EMC came down to three factors: flexibility, reliability and scalability. "We are not sure what applications will continue to stay on the mainframe and which will ultimately migrate to NT. A strength of EMC is its flexible storage reconfiguration."

Duren says that one application, a document imaging system that once resided on a standalone NT server platform, has been migrated onto "the NT-side of the ClearPath mainframe. Its data is stored on the EMC Symmetrix array." Without such application and storage consolidation capabilities, he notes, "we wouldn’t be able to manage everything with only six people."

Says the manager, reliability is "another strength of EMC." He reports that the Symmetrix array automatically contacts EMC when an impending fault is detected, "I got a call from my [technical support representative] informing me that he would be in the next day to replace the disk pack that was going bad. The system reports its own drive failures and other problems. I wasn’t aware that we had a problem until I got the call that the replacement disk was already on its way."

Among the most important criteria for the EMC purchase, says Duren, was the "elbow room" afforded by the platform. "We were about to do Y2K testing and we knew we needed enough room to duplicate a large database, so we purchased more disk than we needed at the outset. Now, we realize we need the storage. Our databases aren’t growing that rapidly, but we are doing a lot more with online technical manuals and imaging – lots of graphics and text. That is where I see the biggest area for growth." The Symmetrix array provides more than enough room to grow.

Rather than crediting his foresight in the purchase, Duren concedes that the EMC platform that Simplicity ultimately deployed was not his first choice, "We acquired a different EMC platform than the one we originally specified. That’s a sign of how fast everything is changing. I’m comfortable that, between the Unisys ClearPath architecture and the EMC storage platform that we have in place today, we are positioned for the future needs of our 150 users."

Storage Centralization and SANs

Duren says that, like most IT professionals, he is watching the development of storage area networks with interest. Storage area networks (SANs), the current buzz phrase in the storage industry, refers to an evolving architecture for separating storage from server-captive configurations and centralizing it within a high-speed switch fabric.

The term SAN is often used in connection with high-concept phrases, such as "virtual storage pool" or "information utility." These futuristic storage architectures, expected within the next five to 10 years, depending on the analyst, will respond automatically with necessary storage resources to meet the demands of applications, servers and client systems. SANs are the enabling technology for this coming storage nirvana.

Within a shorter timeframe, some estimate by 2001, early SANs are expected to begin appearing to address the hidden cost of storage: management. For the past decade, SAN advocates say, inexpensive disk technology and increasing capacities per disk have encouraged companies to buy storage in unprecedented amounts. Analysts estimate that, for every dollar spent on storage, between $11 and $18 must be spent annually for storage management. Simple math suggests that, with the storage acquisitions of large companies approaching the petabyte range, some sort of technology fix will be required to prevent storage management costs from breaking the IT budget in many companies. Most analysts expect SANs to provide that fix.

SANs provide a means for centralizing storage so that it can be readily managed by fewer personnel. The technology derives in part from mainframe direct access storage devices (DASD) interconnected via high-speed fiber, such as IBM’s ESCON. Modern SANs use gigabit-speed Fibre Channel and specialized switch products from vendors to provide a robust, centralized storage fabric. When fully realized, these SANs will provide heterogeneous servers with access to a common storage platform. The storage area network will serve as a back-end I/O network to servers and other platforms, which are interconnected via local area networks (LANs).

Unisys storage partners are hot on the trail of SANs, though both companies are conservative in their estimates about the timeframe for the realization of truly "open" SANs.

EMC, for example, has built a SAN-like offering around its Symmetrix storage array products, calling the architecture an enterprise storage network (ESN). Introduced in November 1998, the ESN architecture leverages existing and developing intelligence in EMC storage array controllers to deliver an enterprise-class centralized storage solution. In the future, company spokespersons say, EMC will focus on building into the solution self-tuning capabilities, performance monitoring and enhanced vendor alert capabilities to enable proactive, reduced-cost management.

The company has also founded a Fibre Channel Alliance with a number of software and SAN switch vendors and has stepped up its interoperability testing with the objective of delivering fully-tested solutions whether or not SAN standards have matured. Says Jim Rothnie, EMC’s Senior Vice President and Chief Marketing Technical Officer, "Vital elements of true enterprise storage networks include platform-independent storage systems, centralized management software, tested interoperability, robust online storage-to-storage data movement, storage-based security and data center class support." He notes that discussions of SANs are currently surrounded by "a lot of visioneering" and states that EMC is delivering SAN solutions that work today.

Todd Gresham, Vice President of World Wide OEM Sales with CLARiiON, prefers not to discuss EMC claims, "We have not positioned our storage products for mainframe data exchange. We compliment Unisys’ direction in the open systems arena and support both UNIX and NT storage."

Gresham says that CLARiiON arrays are designed to provide scalable, centralized storage for open systems platforms, "We offer advanced end-to-end Fibre Channel products that are price competitive with other storage array products available today. We are steadily adding features and functions and have terabytes of storage in SAN configurations right now as part of our SAN interoperability and quality assurance testing program."

CLARiiON’s array products, according to Gresham, feature a disk processor enclosure (DPE) that provides dual controllers with fail-over capability and a disk array enclosure (DAE) that can be scaled to over 110 disks controlled by the DPE. The architecture, says the vice president, provides many of the benefits of future SANs today. "We tell our customers that when they deploy our product, they are deploying the plumbing for the open SANs of tomorrow."

EMC, CLARiiON and other intelligent storage array manufacturers regard their products as "SANs in a box." The external array represents the first effort to detach storage from server captivity. "It has typically been the case that, as the price of the server increased, storage was more likely to be externalized," says Eric Herzog, Senior Director for Market Development for Mylex Corporation (Fremont, Calif.), a supplier of internal and external storage controllers for Unisys. "What is significant is that there is now an increasing tendency to externalize storage even in the Intel space."

Herzog explains that externalized storage offers more sophisticated features than internal, server-captive storage, "including more host and drive connections, mirrored caching, cache coherency schemes, volume mapping and other features that increase their price but also provide support for fault tolerant configurations. With our failover/failback controller technology, redundant paths are available to servers using the storage. That is a benefit over other storage controller configurations."

Obstacles to SANs

Herzog says that, despite steady advances in controller technology, considerable work needs to be done before SANs will be fully realized, "For one thing, there needs to be intelligence on the controller that understands the requirements of an application in terms of short or long blocks. There are nuances to various applications and the RAID levels they require."

Veritas Software’s (Mountain View, Calif.) Senior Director of Product Management Mark Griffiths agrees with this assessment and adds that there are several other SAN requirements that the market has yet to address: "Today, a volume or disk set is managed by a single host machine. [To share disk resources in an array or a SAN], you need to create the capability for multiple hosts to manage volume and disk."

Griffiths refers to the volume management issue that continues to plague early SANs. The problem is that disk volumes have traditionally been formatted, partitioned and accorded fixed sizes by the server operating system. To add a larger capacity disk volume to a server, the server needed to be taken offline, data needed to be backed off the existing disk, a new, larger disk had to be installed and formatted, and then backed up data had to be copied onto the new disk. This process created substantial downtime and also limited the capability of multiple hosts to use the volume natively.

Some critics of SANs argue that the concept of a volume must first be divorced entirely from the server operating system before true SANs can be realized. Neither UNIX vendors nor Microsoft show any inclination to facilitate this requirement.

At present, the only solution to the dilemma of volume sizing and sharing is a workaround from Veritas, says Griffiths. The company released an NT Server compatible version of its Volume Manager product in June that, Griffiths notes, "has been shipping on Solaris and SCO UNIX for four or five years." The product enables a volume to be resized without going through a laborious operating system-dictated procedure. The solution does not, however, address the need to share volumes between multiple, heterogeneous hosts.

Storage management also represents a challenge for SANs, Griffiths acknowledges. At present, the monitoring of storage device function and status must be accomplished on an out-of-band basis. That is, LAN-based rather than SAN-based software is used to perform whatever monitoring and control functions are available with the devices. The Fibre Channel Loop Community is working on an in-band management mechanism (one that will use signaling across the SAN interconnect fabric itself). However, the effort faces substantial resistance from vendors of proprietary storage and server hardware.

Another obstacle to the realization of SANs, according to Griffiths, is the lack of SAN skills, "There is no SAN training yet. No one is really implementing them yet. There is a high demand for professional services to assist with SAN deployments, but few organizations possess the combined skills in storage and networking that are required to understand SANs."

Whither Centralization?

Centralization of storage, whether via external disk arrays or SANs, is clearly a growing trend in the storage field. The trend promises to gain momentum as the cost of storage management mounts. Centralized array platforms are enjoying increased attention as centralized storage solutions that are available today. However, IT planners need to keep their attention trained on the future of storage platforms as well.

SANs offer a value proposition that, when realized, will literally change everything. When SANs arrive, server vendors will see a substantial drop in revenues as storage devices are detached from servers entirely. Do the math: Upwards of 60 percent of server pricing is typically associated with installed storage devices in the server box. With SANs, the storage component can be separated from the box.

SANs are also likely to dent the revenues of large-scale proprietary disk array manufacturers. With a fabric-based SAN, it doesn’t matter whether the storage products behind the SAN switch are name brand or off brand – provided that they comply with SAN management protocol standards. Array manufacturers know this and are either positioning products to play in this changing market or are stubbornly trying to keep customers locked into proprietary products with warranty locks and other mechanisms. There are also signs that some vendors are deliberately interfering with standards-making efforts to push back the arrival of open SANs – a technology regarded by some analysts as "the great leveler."

At the same time as centralized storage advocates inch forward toward their definition of storage nirvana, another class of storage solutions – Network Attached Storage (NAS) – is being increasingly seen in corporations. The premise of the NAS device is to place storage on the LAN wherever it is needed by space-hungry end user workgroups – in short, to decentralize storage.

NAS has the advantage of harnessing a thin server operating system kernel optimized for I/O functions and network attachment. The Unisys FS 1006 (formerly NAS2000), as well as offerings from Procom Technology (Santa Ana, Calif.), Network Appliance (Sunnyvale, Calif.), Auspex Systems (Santa Clara, Calif.), and others, offer an ease-of-installation-and-use value proposition that promises a tactical fix to storage requirements. Most analysts believe that NAS systems will eventually provide LAN-based portals to SAN-based storage.

Most industry watchers agree that the cost to companies from Y2K bugs will be quickly dwarfed by the "discovery" within many corporations of the true cost of storage management. Centralization strategies with enhanced management automation are a prerequisite for preventing the storage explosion of the 1990s from becoming the IT disaster of the 2000s.

About the Author: Jon William Toigo is an independent writer and consultant specializing in business automation solutions. He is the author of eight books, including The Holy Grail of Data Storage Management and Disaster Recovery Planning, 2nd Edition, both from Prentice Hall. He can be reached via e-mail at jtoigo@intnet.net, or through his Web site at www.toigoproductions.com.