isham research
On 4 June 2002, IBM and Hitachi announced a comprehensive agreement. Its main provisions were:
The precursor agreement was signed a year before and made each company's storage interfaces, features and intellectual property available to the other. It has worked well in practice, and in many ways the new agreement is a highly logical extension.
It takes things a lot further. The enterprise storage market has evolved - during 2001 Hitachi Data Systems increased its shipments from 9.8 petabytes to 18.1 petabytes - overall growth of 86.4%, although revenue grew much less quickly. IBM shipped its 10,000th Shark storage subsystem to Commerzbank in Frankfurt, Germany. Both HDS and IBM are listed in the 2002 Red Herring Top 100 and are among four or five companies singled out for special attention:
Several recent advances, including IBM's "pixie dust"--which allows the grains on which bits of data are stored to be much smaller--have raised the density limit. IBM and innovative firms, like Hitachi and Seagate, will be the ones that continue to up the ante.And this was even before IBM's 11 June 2002 announcement of 'Millipede' - a thermo-mechanical capable of multiplying current storage densities. In contrast EMC - once a company that took the whole storage sector by the scruff of the neck and changed its direction - is struggling to adjust after 1Q2002 revenues fell to $1.3 billion from $2.3 billion in the same quarter a year ago. EMC's software revenues were hit just as hard as hardware revenues - a drop from from $467 million to $282 million - a worrying sign for a company that claims software will be its future strength and will inevitably need to invest.
The 2001 deal made IBM's and Hitachi's proprietary interfaces available to both partners. This greatly increased their relevance to the market, allowing independent software vendors to develop for both at the same time and also opening up a vista of a future common architecture. They are continuing to compete, as evidenced by Hitachi's announcement of its 9900V range and IBM's announcement of the third generation of Shark, codenamed 'Silvertip'. This competition will continue, even if EMC becomes a spent force.
The first agreement was an early confirmation of a major strategy change at IBM - coming shortly after the decision that virtualisation - a vital component of future enterprise storage solutions - belonged within an overall concept rather than inside proprietary systems such as IBM's Shark. This decision also helped to pave the way for the later hardware/software responsibility split. Shortly before the most recent announcement Hitachi had also previewed its new 9900 V products - which contain so-called 'virtualisation assists' - and its future adoption of the Storage Networking Industry Association's Common Information Model (CIM. IBM followed up with its own commitments to SNIA standards in a detailed storage software 'roadmap' and the announcement of "Bluefin" - a specification designed to use the SNIA's CIM and Web Based Enterprise Management technologies (WBEM) to discover and manage resources in a multi-vendor SAN through common interfaces.
Although the joint venture will initially only take over hard drive R&D and production, other ideas are being discussed:
In addition, IBM and Hitachi are proceeding with separate negotiations related to a planned multi-year alliance to research and develop new open standards-based technologies specific to next-generation storage networks, systems and solutions. The companies have created a process to review joint projects designed to improve interoperability, reduce complexity and improve cost of ownership for storage systems customers. Additional information on this alliance will be announced when negotiations are finalized.
Virtualisation permits a common storage function to act as a backing store and repository for all of an organisation's data. IBM has traditionally been regarded as the world's most capable supplier at the corporate level, delivering solutions to the largest users in the world. Its manufacturing operations, however good their technology, are inevitably being hit by the price wars raging in consumer electronics - the low margins now achievable in electronic manufacturing stand in sharp contrast to the more profitable software and services sectors. Against this background, IBM's decision to form a separate Storage Software Group - headed by recent recruit Mike Zisman - within the Storage Systems Group was a clear preparation for IBM and Hitachi to assign these functions to those best equipped to deal with each. This was a major strategic shift for IBM - continued with the acquisition of Metamerge, a directory integration company that had come to Zisman's attention when he was at Soft*Switch. IBM has also recently disposed of other low-margin manufacturing operations such as Mylex - its low-end RAID operation - and Endicott, which many older IBMers regarded as the spiritual home of the company.
The choice of Hitachi as an in-depth technology partner is no surprise. IBM and Hitachi have co-operated for a long time, with the first major result being IBM's 3900 laser printers. Over the years, co-operation has become closer and closer - Hitachi built its Pilot series using IBM processor chips, and IBM's latest z800 mainframe is assembled by Hitachi using IBM technology. In return, Hitachi has the right to relabel the product for the Japanese domstic market. Hitachi also has an impressive reputation as a storage manufacturer dating back to the early 1980s - making it a natural choice as a hardware R&D and manufacturing partner.
IBM will now be free to concentrate resources - investment dollars and management time - on realising the Storage Tank concept first discussed almost two years ago. In principle Storage Tank has the potential to become as successful as the original System/360 - a universal system for all enterprise storage needs. While Storage Tank has a sound strategic architecture, it is a major project and realising its potential will require significant investment; perhaps the lion's share of IBM's storage R&D budget.
There are other benefits for both IBM and its customers - the emphasis on support for heterogeneity in Storage Tank makes it a cross-platform solution, and this is relevant both in customers' existing multi-vendor situations and across IBM's server platforms. IBM has committed to making source code references available, publishing the interfaces, and basing the system on Linux - a marked contrast with previous products such as Shark, which were based on highly proprietary AIX variants. IBM can also combine budgets that would otherwise have served only one of its platforms into an effort that serves them all, and customers can commit to storage investment without regard for future platform decisions.
In a sense, the realignment now taking place among the niche SAN/NAS vendors has parallels in the early mainframe days. The IBM System/360 was announced with some thirteen storage products, each tailored to a price/performance tradeoff level within the storage hierarchy. It looked fine on paper - but no one exploited it because of its inflexibilty in response to business need changes. The ingenuity of system designers led to them finding ways to avoid using the expensive - in cost/bit terms - devices such as drums and head-per-track disks. By the early 1980s IBM was marketing only one mainframe storage device - the 3350, succeeded by the 3380 - and paying lip service to the storage hierarchy concept by occasionally selling a short string. "Balanced systems" replaced the storage hierarchy concept, and hardly anyone noticed.
The same is now happening in the SAN/NAS environment. Vendors of exotica - many of them financially extended or still burning venture capital - are finding that users design around the requirement for their products; bandwidth and capacity are much cheaper than expected and designs correspondingly simpler. Exploitation of raw technology is becoming less important - the concept is starting to count. Storage Tank also represents a convergence of SAN and NAS technologies - most solutions even now use elements of both.
One area where the joint venture will dominate is intellectual property. IBM regularly tops the patents league tables with around three thousand new patents each year - 400 or so directly related to storage. Hitachi was 8th in the 2001 patent league tables with 1,271 new patents. At times both have also increased their portfolios by acquiring the patent rights of others - when National Advanced Systems sold ISS in the 1980s it retained the patent rights, which then passed to Hitachi Data Systems. EMC has also played this game; before selling Data General it floated a separate company - DG Patent Holdings Inc - to hold the patent rights and keep them out of the scope of reciprocal agreements it had with IBM. The IBM-Hitachi joint venture will take over both IBM's and Hitachi's hard drive-related patent portfolios. This transfer could effectively extinguish EMC's patent violation suit against Hitachi. Similarly, the joint venture's possession of manufacturing facilities within the United States would render the ITC investigation requested by EMC at the same time irrelevant.
Among other things (I've Been Misled, Insert Bug under Mask, and International Brotherhood of Mechanics) it's often suggested that IBM stands for "I've Been Moved" - a reflection of the functional rotation that occurs within the company. In the past this has been a problem - most of IBM's competitors have nothing like the range of products that IBM has even within one of its divisions. Few IBM managers and executives have the opportunity to focus on one activity for the bulk of their careers, yet their competitors are specialists - EMC sells only storage. One advantage the new joint venture will have is clear direction; its first chief executive - Dr Jun Naruse of Hitachi - is a career disk engineer with an impressive record of achievements such as the first sealed HDA disk drive, marketed as the NAS 7360.
With the two competitors that gave it the most trouble last year now even better organised, EMC's situation is as bleak as it has ever been. None of the issues highlighted by industry observers and analysts last year have been addressed. Despite all the comments about IBM's Storage Tank being late - three years from first discussion to delivery is another parallel with System/360 - it is a cohesive strategy with open commitments that users and other vendors can build on. EMC's software strategy looks opportunist, with the company apparently undecided which way to jump. A strategy that includes buying-in innovative software is all very well if the owners want to sell - but it's hard to describe your overall strategy if you haven't found the components yet. A merger with BMC - rumoured to be undergoing due diligence - would give EMC access to BMC's Patrol, but might have interesting accounting implications because the companies treat their software revenues very differently.
The strategy of buying hard drives from the lowest bidder and wrapping availability enhancements such as RAID around them - a 'box full of discs' - starts to fail as the number of manufacturers selling at the factory gate dwindles and conceptual functionality becomes more important. As far as the ageing Symmetrix is concerned, Fujitsu still needs a route to market for its high-end storage systems - marketing via the Amdahl services organisation is proving extremely ineffective. Even if either deal comes to fruition, the resulting partnership will be at least two years behind IBM and Hitachi in the realisation of any concept.
Users making storage investment decisions should look to the medium term rather than the short, and the long term rather than the medium. IBM's commitment to openness makes hardware decisions relatively unimportant and price/performance will be key - but the selection of a long-term storage strategy should be approached with care.
© July 2002 isham research
isham research Home Page and Contact Information