In-Depth

The Ever-Changing Microsoft Repository

Microsoft Corp. has quietly altered its repository strategy after learning what many others had discovered earlier — that a cost-effective repository supporting multivendor development tools is too complex for today’s technologies. The company now plans to utilize multiple repositories, each built for specific applications.

Observers repeatedly point to IBM’s AD/Cycle fiasco of the late 1980s as proof that a development panacea utilizing a single reposi-tory to make all code available to each of a project’s developers, at any time, is still little more than a dream. While Microsoft has faced numerous hurdles in its foray into repositories, its setbacks have been remarkably far less public than IBM’s failures. But unlike IBM, which halted work on its repository when it was apparent it would not work, Microsoft has continued to evolve its effort in response to setbacks.

In the latest Microsoft strategy, the so-called Microsoft Repository — under development since mid-1997 with Platinum Technologies/ Computer Associates — is now aimed at data warehousing projects rather than development projects. Microsoft officials say development of a new data repository for developers is underway in Microsoft laboratories. Observers note that Microsoft has kept its recent plans close to the vest, following criticism in the past from some quarters that the proposed repository was a solution in search of a problem.

“There have been a lot of [changes] in how people look at repositories over the past couple of years,” said Steve Murchie, group product manager for data warehousing at Microsoft. “A couple of years ago, we started looking at how we could integrate meta data, which began the change [in Microsoft Repository strategy] from its former life as a development repository. We have done a lot of work with vendors in the data warehousing business to find common ways to share meta data.”

He added that the emergence of the eXtensible Markup Language (XML) has helped shape the new Microsoft repository efforts. “XML is the key technology for us now,” Murchie said. “It’s well suited to working with different types of meta data.”

Microsoft shipped Version 2.0 of the repository this spring as an integrated component of the SQL Server 7.0 database, forming the foundation of the company’s Data Warehousing Framework. The latest repository is still shipped with the Microsoft Visual Studio and Visual Basic toolsets, though that will change once the new development repository is completed, Murchie said.

Mike Budd, an analyst at London-based consulting firm Ovum Ltd., added, “Version 2 is better than Version 1, but there’s still not enough performance and not enough basic services for tool vendors to support it.”

Shifting strategies

Analysts express little surprise that Microsoft has shifted its strategy several times since announcing plans for the repository a few years ago, noting that many firms have been forced to abandon plans for enterprise repositories through the years. “The whole repository thing has proceeded at a truly glacial pace,” noted Lou Agosta, an analyst at Giga Information Group, Cambridge, Mass.

“A problem with repositories is that they’re quite a difficult technology to build,” said Ian Bramley, chairman of the Enterprise NT Management Forum at Butler Group Ltd., a consulting firm in East Yorkshire, U.K. “You can count the successful ones — UREP [from Unisys Corp.] and Composer [recently acquired by Computer Associates International Inc.] — on one hand.” He also noted that as software development evolves — from host-based to client/server and the Web — repository development efforts must be updated as well, a very difficult and time-consuming chore. Other analysts note that in addition to technical issues, Microsoft’s repository team has faced skepticism internally from the Microsoft tools group, which must support the repository group in order for the plan to work.

Despite the rocky pace of the Microsoft repository effort, observers give the Redmond, Wash., giant generally high marks for what some call a sensible approach of changing strategies when necessary rather than abandoning the effort. Microsoft is also applauded by other experts for its apparently strong role in the creation of meta data standards by the Meta Data Coalition, a standards consortium of data warehouse vendors and users based in Austin, Texas.

“Microsoft is still an active member, but not as involved as the eight or 10 lead companies,” said David Marshall, chief technical officer at Evolutionary Technologies Inc., Austin, Texas, and a member of the standards development team.

Microsoft moved to promote the meta data standards effort following its development in 1997 of the Open Information Model (OIM), a vendor-neutral meta data standard based on the Unified Modeling Language (UML). The effort included input and review from more than 20 tool and data warehouse software suppliers, including Rational Software, Logic Works, Texas Instruments, Select Software Tools, Popkin Software, Business Objects, Cognos, Informatica, Prism and Platinum Technologies. Microsoft turned over all rights to the OIM to the Meta Data Coalition at the end of 1998. Microsoft officially joined the Meta Data Coalition at the end of 1998.

On the other hand, observers say Microsoft has been slow to actively support connectivity between the Open Information Model and the MetaObject Facility (MOF) meta data standard built by the Object Management Group, Framingham, Mass. The Universal Repository (UREP) from Unisys Corp., Blue Bell, Pa., is the basis for the OMG’s MOF specification. Unisys executive Sridhar Iyengar, who is currently the liaison between the OMG and Meta Data Coalition, oversees the effort to link the MOF and OIM specifications.

“Microsoft could have done more to bring the groups together,” said Iyengar, a Unisys fellow who heads up the meta data and object architecture units at Unisys. “They [Microsoft] have taken a more active role in the OMG process over the last year. There has been some progress over the last year, but not enough.”

Original plan

Development of the Microsoft Repository began in 1994 under a joint development agreement signed by Microsoft and initial repository partner Texas Instruments Inc. With great fanfare at the time, the two organizations outlined plans to build a repository that would support Microsoft’s Windows and Windows NT platforms, as well as Unix and IBM’s AS/400 and MVS operating systems. Three years later, unhappy with the progress of the project, Microsoft switched development partners, and signed an agreement with Platinum Technologies to take over the partnership and the rights to sell the repository on non-Microsoft operating system platforms.

Ironically, both Platinum and Texas Instruments’ software development group are now part of Computer Associates International Inc., Islandia, N.Y., which is continuing the joint repository development with Microsoft. CA acquired Platinum and its repository technology in 1999. And this spring, CA completed its acquisition of Dallas-based Sterling Software, which had acquired the Texas Instruments software business just before the scuttling of the joint development pact with Microsoft in mid-1997. Under the current agreement, CA retains exclusive rights to sell the Microsoft Repository on MVS, AS/400 and multiple Unix-based systems, as well as on Windows NT systems running non-Microsoft database management systems.

Microsoft’s initial repository strategy called for building a system to store and share data created and used by software development teams working on one or more projects. As the project grew, Microsoft pushed the repository as a meta data management solution for a variety of applications, including business process reengineering, data modeling and design, packaged applications, legacy application maintenance, data warehousing, knowledge management, tool integration and, of course, application development.

Microsoft officials admit the plan was similar to IBM’s costly AD/Cycle effort that failed miserably more than a decade ago. In that case, IBM and several partners were building a repository that promised to support a variety of CASE tools from then-industry leaders like KnowledgeWare, Index Technology, Bachman Information Systems and other suppliers whose stars fizzled quickly following the demise of the AD/Cycle program. Observers note that IBM’s plan to build a repository for developing mainframe COBOL applications was finishing just as the IT world began shifting to the client/server model.

Microsoft officials concede that many of the same lessons were learned during its own repository development endeavors; this led to the shift to focus the effort on data warehousing instead of on application development. “In data warehousing, there are fewer transactions, but they are larger in scope [compared to software development projects],” said Microsoft’s Murchie. The repository, now shipped as part of the SQL Server database, was far better suited for data warehousing applications, Microsoft officials decided.

Multiple repositories

“We decided that there are ‘big R’ repos-itories and ‘little R’ repositories,” said Murchie. About two years ago, Microsoft officials “started to see why the movement to a ‘capital R’ repository, one that is built for all of the applications, has never succeeded. It was not a workable solution. On the development side, we felt that we had to look at a slightly different architecture that wasn’t built on the traditional Microsoft Repository architecture,” he said.

At some point during the last year, Microsoft therefore shifted its strategy to embrace multiple repositories, said Murchie. He declined to discuss details of the new repository except to say it will incorporate XML technology and will support existing repositories, including Unisys’ UREP and the Platinum Repository. The development repository will also be based on Microsoft’s SQL Server database and will be built by Microsoft engineers without help from outside partners, Murchie said. He added that the new system would be unveiled in the next few months.

Budd at Ovum said the multiple repository plan makes sense only if each of the systems shares a common meta data store. “There is data that needs to be shared between the data warehousing and development environments,” he said. “If there is not a common meta data store, XML may work in the future, but not as it is today.”

Murchie noted that other repository suppliers, specifically Computer Asso-ciates, have also shifted to a strategy of offering multiple repositories. “That’s what [corporate] users want,” he said. “People are tending to have a lot of ‘little R’ repositories.”

Alan Houpt, senior vice president of business development at CA, said that “the Platinum Repository is available in a lot of different versions” for different applications. Each of those versions will soon incorporate the Microsoft Repository technology, which is now shipped as a standalone system for non-Micro- soft platforms. The Platinum Repository is a result of the marriage of the two primary IBM Repository competitors, Brownstone and Reltech, which were both acquired by the former Platinum to facilitate the integration of the hundreds of tools it acquired during a buying spree in the mid-1990s. Microsoft officials say the advanced technologies in Platinum’s combined Brownstone and Reltech repository prompted its decision to abandon Texas Instruments as a development partner in favor of Platinum.

Bank of America, Charlotte, N.C., saw the need for a repository to organize the data and meta data in its repository several years ago, just as Microsoft was beginning its Microsoft Repository project. Five years ago, two years into a corporate data warehousing project, the Bank of America IT unit found it “difficult to keep up with everything in the warehouse,” said Craig Bell, vice president of meta data resource management at the bank. “There was too much data, too much meta data and too many users,” he said.

The Microsoft partnership had some people concerned about whether the repository would continue to support and improve the mainframe and DB2 database versions used widely at the bank. “Whatever happened, we wanted to make sure it would run on DB2,” Bell said. “The company said they would continue to support DB2 and have kept their word.”

Bell said his unit is interested in the Microsoft Repository, but as it evolves, “we’re still in a wait-and-see mode. We’re definitely keeping an interested eye on what’s going on.” He said Bank of America will be seriously interested once the Microsoft system supports DB2.