In-Depth

What's behind BEA's big bet on tools?

• By Jack Vaughan
• August 1, 2002

San Jose, Calif.-based BEA appears to sense it may be time to turn its back on the notion that tools should be completely independent of runtime software, and it is now ready to push tools along with its platforms. It is also wagering that Java needs a mainstream IDE to attract wider use. Thus, the firm is beginning to promote its WebLogic Workshop software at the same time it shows diminished interest in technical Java tools from Santa Clara, Calif.-based WebGain, a third-party tool maker BEA helped fund, but which has struggled in recent months.

A number of traits have merged harmoniously, creating success for the WebLogic server over the last few years.The BEA application server was fast, had an architecture that impressed enterprise shops with transaction processing backgrounds, and it adhered carefully to the new J2EE compatibility standard. With the Java application server, the foggy notion of 'middleware' gained definition, and BEA's WebLogic server was the most brilliant example.

It was a genuine hit, driving the company from $61.6 million in revenue for the fiscal year ending January 31, 1997 (the year before WebLogic's acquisition) to $975.9 million in revenue for the fiscal year ended this January 31.

But the fast-moving technology wheel always seems to offer new software poised to unseat established successes, and BEA must now convince customers that its Web services architecture bests or betters others. First estimates place it in the thick of a Web services battle that does not - at least in the Java camp - have a leader. Even as it seems to reach greater accommodation with HP, BEA is encountering renewed competition from the likes of IBM, which, by some estimates, may surpass BEA in the app server market this year; as well as Sun, which recently began offering a free version of its iPlanet app server bundled with its flagship Solaris OS.

BEA is not standing still. The tools and strategies the firm has put in place for its next push forward amount to virtual 'bet-your-business' propositions. But it is a difficult environment in which to execute.

Like all software companies, BEA faces tough economic going. Yet it recently increased R&D spending. The limits to future growth, BEA's managers have decided, are a matter of 'peopleware.' The key to gaining greater share, they wager, is to expand the number of developers capable of deploying distributed enterprise Java systems. In a tools battle that seems to boil down to hardcore Java vs. .NET, BEA is set to provide an interesting alternative.

The emerging BEA strategy uncovers one aspect of the shiny Java language success story that is a bit dark, at least for corporate developer groups. Despite incredible growth, the number of Java developers still lags far behind the ranks of Visual Basic and Cobol developers. It is disturbing that Java implementations continue to increase in complexity, requiring higher and higher degrees of developer skill.

In short, enterprise Java is still something of a specialist game for true computer jocks.

The tool that holds the promise to vastly expand Java use is BEA's WebLogic Workshop, formerly code-named Cajun. The software is just coming to market, about a year after the company acquired its originator, start-up Crossgain Corp. of Redmond, Wash. Highly regarded as a development firm, Crossgain was composed in large part of former Microsoft employees, including such notable Microsoft tools gurus as Tod Nielsen and Adam Bosworth.

BEA and others hope the Crossgain crew can explode out of the gate with the type of developer-friendly software Microsoft is known for.

Clearly, this will be a new era for BEA. Until now, noted Gartner Inc. analyst Yefim Natis, BEA has focused on one major objective: selling application servers.

'They focused on just one goal. They refused to enter the development tool market,' said Natis. 'They also avoided the integration market and the messaging market, despite pressure. They only focused on new-generation online transaction processing.'

BEA succeeded so well, 'because all their wood was behind one arrow,' added Natis. 'But now they are paying for it.' Tools are just part of the new push at BEA. The company has also recently upgraded its messaging, portal and application integration offerings.

Tuxedo junction
Its founders seemed to expect a lot when BEA was established in 1995, but few industry observers shared such high expectations. Founders William Coleman (now chairman) and Alfred Chuang (now CEO) came to the task with established credentials, having served long stints at Sun Microsystems. (Edward Scott, a third founder, worked for Pyramid Technologies and in the U.S. government. Scott now serves BEA as a consultant.)
Coleman's and Chuang's first big move came in 1996, when BEA acquired the former AT&T's Tuxedo transaction processing (TP) monitor group from Novell. If observers expected the new owners to merely milk a proven technology, they could be forgiven - but they turned out to be wrong.

With the purchase, BEA became one of the top 100 independent software firms, garnering a 36% share of the worldwide distributed online transaction processing (OLTP) market. That market might have been described as the high end of client/server computing, then in something of a decline. BEA would be widely deemed the buyer of mature (read: 'somewhat unexciting') properties, and the impression was enforced in 1997 when BEA acquired DECmessageQ message-oriented middleware and ObjectBroker object request broker software from the fast-receding Digital Equipment Corp.
Later, the company would buy NCR Corp.'s Top End transactional middleware product. By 1997, the company had achieved a respectable $61.6 million per year in revenue and gone public. It was a major enterprise middleware player, but that segment was changing as Java came to the fore.

Tengah happens
This string of well-executed mergers somehow set the stage for the greater coup - the September 1998 acquisition of WebLogic Inc., the maker of an application server then known as Tengah (named after a city located in central Java in Indonesia), which was just gaining notice.

BEA's lineage in transaction processing influenced its decision to buy WebLogic, which was far from alone in a diverse Java app server segment, said WebLogic manager Scott Dietzen, who serves today as BEA's CTO.

He is well-known for helping to develop server-side standards within the Java community, but it is less known that Dietzen worked as principal technologist at distributed transaction processor maker Transarc prior to WebLogic. Dietzen confided that he and WebLogic developers approached the problem of building an application server as they would the problem of building a transaction processor. He suggests BEA saw that trait, and looked to exploit it with its Fortune 500 commercial transaction processing customers like FedEx, United Airlines and others.

In conversation, Dietzen will remind you that the first wide interest in Java was as a client or potential Web page animation architecture. He and a few others saw it immediately as the foundation for a server architecture.

'Tuxedo was the high end of client/ server; [that is,] it looked more like mainframe architecture,' Dietzen recalled. 'The Web made that kind of processing architecture useful for everybody - not just a FedEx doing more than 10,000 requests per second. Now WebLogic is a bigger product than Tuxedo ever was,' he commented.

'WebLogic was a transaction processor. It was very well architected. We had business-critical production in 1996. That was the reason BEA picked it,' he said.

He added that the WebLogic app server 'had the reputation that it worked,' and worked at a clip that satisfied newly converted C++ developers. 'It was written in Java, but it was fast,' Dietzen said.

BEA deftly capitalized on this new acquisition. It kept the WebLogic team in place, but supplemented it with long-time transaction hands. 'We put core BEA Tuxedo gurus to work on hardening it. That gives it a different [lineage] than other app servers,' explained Dietzen.

Said Eric Stahl, director of product marketing for WebLogic Server: 'BEA's experience with Tuxedo made the move into the application server space a natural fit. The vast amount of functionality [he points to load balancing, failover and caching] is conceptually the same.'

Stahl also touched on a theme that resonates through several BEA purchases. 'BEA buys companies to get people, not products. While Tengah was a great start, it was [about] getting the team in which we were most interested.'

Gartner's Natis also provides a take on how WebLogic became so successful. 'BEA came into the market with a good reputation for Tuxedo, at a time when Java was suspect and not very proven in the enterprise. BEA came into the picture of Java computing as an expert for the enterprise with an established base,' he said.

And BEA benefited from gaffes among larger competitors, said Natis, who noted that IBM's app server was often 'bought but not deployed until it reached Version 3.45.' Both Oracle and Sun took missteps, fielding a variety of server versions that were sometimes discontinued.

'For a while, BEA had very incompetent competition,' Natis said.

Just as competitors continue to work to better their engines, BEA is at work on improving WebLogic. The measure of the work, as happened earlier in the database server segment, is a stream of benchmark ratings for every new release. For its part, BEA recently reported ECperf benchmark results for WebLogic Server 7.0 running on the HP 9000 rp8400 server that show the system was able to process 37,791 Bbops/min@Std (business operations per minute), and claimed a new ECperf world performance record. ECperf is a standard for measuring the performance of Java 2 Enterprise Edition (J2EE) servers.

Although it is often positioned as a high-end offering, BEA's Dietzen is quick to point out that WebLogic's ECperf figures show a favorable total cost of ownership (TCO). The price/performance result of the ECperf benchmark, representing the TCO of the tested configuration, including all hardware, software and database components, was $38/Bbops/min@Std.

Such performance has translated into leading, but challenged, market share for BEA. Gartner, Meta Group, IDC and AMR are just some of the research groups that put BEA on top of the application server world, followed usually by IBM, Oracle and Sun. IDC, for its part, says the market will double to $4.4 billion by 2006. Estimates like that ensure BEA will remain in the gun sights of the biggest software players.

IBM's lucrative mainframe, integration and messaging software businesses continue to provide inroads for the IBM WebSphere server brand. Several of the analyst groups see the gap between BEA and IBM closing, while some project IBM will surpass BEA in app servers this year.

BEA's Dietzen is not alone in asserting that IBM's WebSphere sales are tied in with many big services deals, that Oracle is looking to sell databases or application suites, and that Sun is looking to sell hardware. BEA app servers run on many of the Solaris boxes Sun sells, so the relationship with Sun remains one of 'co-opitition.' BEA has endeavored to forge deals with IBM's services competitors such as Accenture, EDS and CSC, in an effort to counter the Armonk, N.Y., software and services giant.

At the same time, BEA has launched its Workshop effort to ease the cost and burden of integration. And it has sought to find the type of software bundling that will have broad appeal. BEA's new WebLogic Platform 7.0 combines the core app server with development, portal and integration frameworks so mainstream application developers, not just J2EE jocks, can quickly build Web services.