IBM didn't call it "Shark" for nothing. Big Blue’s new Enterprise Storage Server (ESS), developed under the code name Shark, is clearly intended to eat the competition alive in the market for big, terabyte-scale hard disk arrays, commonly referred to as RAIDs. But there are questions about the ESS's ability to handle "virtual" storage—questions that suggest IBM may be more interested in market dominance than in giving IT managers what they need.
Since 1994, IBM has OEM'ed a high-capacity RAID subsystem from Storage Technology Corp. (STK). Called the RAMAC Virtual Array, it’s enabled IBM to win approximately 6,000 customers—about one-fourth of the total market for such systems. STK itself also brands and sells the same device. However, its OEM agreement with IBM expires at the end of next year, and obviously IBM does not intend to renew it.
Within IBM (where they say "ESS," not "Shark"), Ron Kilpatrick, general manager of IBM's Storage Systems Division, has been telling analysts that "the ESS is designed to grow with customers' needs and to easily incorporate the latest storage innovations. It incorporates a snap-in design that allows users to add capacity, performance, and connectivity over time using the same platform. This design allows customers to evolve their storage systems well into the future, making the ESS particularly well-suited to server and storage consolidation activities and offering unparalleled investment protection."
The ESS is specified to be scalable from 420 GB to at least 11 TB, with two four-way symmetric multiprocessors (SMPs) on board and IBM's proprietary Serial Storage Architecture (SSA) linking the drives internally. Battery backup renders the cache memory essentially nonvolatile. End-user pricing, when the ESS is released later this year, will probably be around 50 cents per megabyte.
IBM is also specifying that the ESS will be plug-compatible not only with its S/390 mainframes and AS/400 midrange systems, but also with ordinary UNIX and Windows NT servers. The ESS will be capable of communicating with other storage devices presently connected with ESCON, Fibre Channel, or Ultra SCSI cabling. "This combination of performance, openness, and flexibility is unprecedented in the world of high-end disk storage systems," Kilpatrick said—and it's certainly true for IBM's offerings, if not for the industry as a whole.
Trying to regain market share
In this niche, IBM's two other competitors are Hitachi Data Systems, whose customer base is mainly in Japan, although it recently inked a deal to supply high-end storage products to Hewlett-Packard Co.; and EMC Corp ., which is the market leader. With the ESS, IBM puts them both on notice that it intends to have the full-meal deal and isn't going to share it—not even with STK.
"Quite simply," Kilpatrick said, "no other vendor has IBM's breadth of storage expertise in disk and tape subsystems, storage management software, hard disk drive technology, systems integration, and support services. Many competitors in the storage industry have only one card to play: They are disk storage companies, tape companies, or software companies. No other vendor can bring together such a full suite of hardware and software solutions to market—backed by the world's largest service and support organization."
Well, the reason nobody (not even IBM) makes too many different kinds of storage products is because manufacturing and supporting them all is too expensive. To succeed in the storage business, you have to specialize.
STK, for example, is the dominant player in the high end of the tape-library market. On the same day as IBM's ESS announcement, STK announced a campaign called "Step Up" to enhance RAMAC installations, regardless of whether those units are branded by IBM or STK. "The 'Step Up' campaign offers functional and performance enhancements for existing customers to purchase the latest virtual disk technology in StorageTek's Shared Virtual Array disk subsystems," said Victor Perez, executive vice president and COO of STK.
As for EMC, it's the market leader in high-capacity HDDs and RAIDs—a market that IBM essentially walked away from in the 1980s. According to market research compiled by Dataquest, EMC's revenues have remained consistently high over the past three years, while IBM's revenues from its mainframe storage business have actually declined. Today, EMC enjoys a market share of almost 50 percent, while IBM holds barely 30 percent.
That "virtual" question
IT managers may or may not care about the vendors' "horse race," but if they're concerned about keeping their enterprise on the leading edge of this technology, they should consider posing a question to their IBM rep: "Is the ESS 'virtual,' in the way that the RAMAC Virtual Array is 'virtual?’" That is, can it separate “logical” from “physical” storage?
With a RAMAC subsystem, managers typically do not need to concern themselves with such tasks as optimizing the usage of disks or volumes. According to STK's Perez, "This enables them to focus on increasing business benefits instead of managing the details of data placement. Virtual storage facilitates efficient use of disk space by handling storage management issues such as disk tuning, workload balancing, and space optimization 'behind the scenes.'"
IBM's ESS will not have such virtual storage capabilities, at least not right away. It's not specified in IBM's printed materials, and it wasn't mentioned or alluded to by Kilpatrick or other spokespersons, except to say that the capability was in development.
"IBM’s Shark offering is being released without virtual technology," one skeptical analyst told me flatly, though not for attribution. "And this is curious," he added, "since IBM captured more than a quarter of the market by capitalizing on the 'virtual' benefits of the RAMAC."
Considering how much STK's RAMAC has contributed to IBM's bottom line over the past five years, the ESS announcement is hard to see as anything but a kick in the rear. My own interpretation of this dustup is that STK's virtual storage technology—arguably the key feature of a RAMAC subsystem—is proprietary to STK, and the OEM agreement under which IBM has been selling RAMAC hardware does not also entitle IBM to license it. IBM is surely capable of developing and manufacturing practically anything its officers might desire; but apparently it can not do so very quickly.
Migrating from RAMAC to ESS apparently means giving up (at least temporarily) those “virtual storage” functions for which RAMAC has been so popular. Ultimately, it's the IT managers themselves who will decide whether they want to stay in the pool with the STK/RAMAC architecture or go out into deep water and swim with the Shark.
The authors and editors have taken care in preparation of the content contained herein, but make no expressed or implied warranty of any kind and assume no responsibility for errors or omissions. No liability is assumed for any damages. Always have a verified backup before making any changes.