The web application server business just got a bit more consolidated, with HP’s acquisition of Bluestone Software, announced during Fall Internet World in New York. As websites begin adding industrial strength commerce and transaction processing the application server has become strategic linchpin. As J2EE standards mature, the application server itself is becoming commodity. There is little room left for niche middleware players. For too long, Bluestone has been the small software company with big technology.
Why have application servers so important to the web, and why have they become strategic application platforms for major e-business hardware and software vendors?
In short, the answer is scalability, availability, and reliability. When an ad appears on television or when B2B trading partners begin collaborating on their supply chains, web commerce sites cannot afford to slow down or fail. And lesson one for any e-business application designer is that, compared to internal applications, those that reach outside the enterprise cannot easily control or predict their user bases and transaction loads.
The logical approach to providing always-on capability is to start distributing databases and business logic on different machines, so capacity can quickly be added wherever and whenever it is needed. In e-commerce, it’s hard to know when you’re going to need another database or application machine. And, in so doing, industrial strength application and database management strategies grow necessary, because loads must be balanced and failover must be applied. While distributed architectures did not start with the web, the web has made them more mission-critical.
Because the web is inherently stateless, the housekeeping functions traditionally performed by the operating system must be performed by something else. Although the database could—and in many cases does—perform services such as persistence and object/relational mapping, the conventional wisdom in high-end web application design is to move that overhead to the application layer, because the database is already pretty busy. Enter the application server. With Oracle 9i, even the venerable database vendor has gotten the message, elevating the application server to the same prominence as its database.
With the emergence of J2EE (Java 2 Enterprise Edition), a framework has emerged that specifies transaction and related services, including object/relational mapping, database transaction persistence, component generation, web page generation, along with APIs to directory services, transaction management, messaging and database access. Today, J2EE is the dominant architecture for web applications that are serious about scalability.
That makes the application server an increasingly strategic buy. Ask most web teams about their short list, BEA WebLogic, IBM WebSphere, and iPlanet (formerly Netscape) Application Server generally make the first cut. Significantly, two of the three are legacy buys—IBM, because of the natural link with mainframe operations; iPlanet, because of its Netscape heritage; and BEA, because they were extremely aggressive in hopping on EJBs before everyone else. What’s also interesting are the once-popular names that are no longer there: NetDynamics, which embraced Java, but not Java Beans, early, and HAHT, whose architecture was C++ based.
Today’s second tier survivors are players that were willing to embrace the J2EE transition and eat their young. SilverStream, which like NetDynamics, embraced Java early, has recently migrated its architecture from proprietary Java components to J2EE; Iona and Gemstone, whose roots were in the ORB and object database fields, respectively, repositioning their technologies as building blocks of J2EE servers.
Last, but not least is Bluestone, which probably made the toughest transition of all. Like HAHT, it had its own proprietary development language (Sapphire), and it was largely C and C++ based. A couple versions ago, Bluestone bit the bullet and made a wholesale migration to a complete Java architecture, rather than take the path of least resistance by “protecting” its legacy base and making only cosmetic changes at the API level. Yes, it would provide migration tools to Sapphire customers, but no, it wouldn’t continue the dated architecture.
It has made some smart technology bets, beginning with its early emphasis on load balancing and failover, which are now taken for granted in the space. Bluestone also hopped the XML bandwagon early, initially with a giveaway XML server product that it eventually incorporated into the core platform. Recently, Bluestone made a shrewd acquisition of Arjuna, a German software firm that developed the first Java-based transaction server, which provides a needed counterweight to BEA’s trove of database-oriented middleware.
The dilemma remained: Bluestone generally got high marks for technology, was effective at getting its message out, but remained stuck in second-tier position in a product category which emerging standards were transforming into a commodity. In short, it faced the usual battle for niche software vendors: swallow quickly or get swallowed up.
HP to the Rescue
Bluestone, which exhibited typical performance for the early stages of a public company, was beginning to show signs that it might finally enter the black. Annual revenues have doubled year over year through 1999, and three quarters in CY and FY 2000, they have already hit the 2x mark, at $28 million. But the company, which focuses on large accounts, has been heavily reliant on its top 10 customers, which last year accounted for over half the revenues. With Enterprise Java technology becoming commodity, it needs the critical mass to convince large accounts that it would be safe to buy Bluestone products.
As the last major software player without an application server strategy, HP Software Division’s acquisition of Bluestone at first glance appears quite logical. HP has recently begun making waves to challenge Sun’s dominance of the e-business space, courtesy of more powerful hardware, tighter relationships with Oracle, a new e-Services initiative, and a president—Carly Fiorina—who is striving to shake the company of its plodding, consensus-driven culture, with a more entrepreneurial focus that is more open to risk taking.
To the uninitiated observer, HP Software sounds like an oxymoron. The company’s tagline, “Invent,” harkens back to a history where clever engineers did neat things devising electronic instruments in the backs of their garages. The company has always had a strong product engineering bias. This author is currently working with one of the fruits of that culture: a 10-year old HP LaserJet II-P which continues to effortlessly turn out 4 pages of high-quality copy every minute, refusing to become obsolete.
But then again, the old AT&T had a similar reputation. Its products were built to last, and with Bell Labs it had state-of-the-art R&D. Chances are, if you took the company up on its offer to buy the phones that it used to rent, prior to the company’s 1984 split-up, chances are the units would still be ringing. Yet, the technology business is littered with the corpses of engineering-driven companies that were tone deaf to marketing, with AT&T the poster child. The Baby Bells spun off from AT&T, which were left with lower margin businesses, have reinvented themselves, with many of them looking far more attractive to investors than Ma Bell herself. Yet, as this report was going to press, AT&T was trying to extricate itself from its latest missteps by once again breaking up the company into business units that might make smarter decisions.
When we look at HP Software, we think of OpenView, a product that invented SMTP network management market. And, in spite of HP’s neglect, OpenView’s Network Node manager has managed to remain the gold standard for network management, even as rivals like CA and BMC begin adding sexier neural network/genetic algorithm technologies to do forward prediction of network operations. This is the same OpenView that withstood the HP hardware company’s embrace of CA’s Unicenter two years ago as the preferred framework for HP-UX platforms.
Since then, HP has formed a software business unit around OpenView. Like IBM Software, HP is striving to form a real software business from a vast array of market-leading and obscure point products. With OpenView providing the pillar into enterprise computing, HP began enunciating its strategy for e-business. the other major pillar of the software: The e-Speak XML framework, and e-Process, for providing high-level tools for designing and managing e-business workflows.
e-Speak, a framework of XML-related, covers XML communications, and the management of XML transaction services. In other words, the concept is sufficiently broad yet vaguely defined that at this point it is difficult to tell whether it will compete with competing frameworks from players such as Microsoft. e-Process is more tangible, currently consisting of ChangeEngine, a tool for developing e-Business processes and workflows, and HP Service Delivery, a telecommunications solution which embeds the WebMethods Enterprise EAI engine.
Bluestone fills a key gap in HP’s e-business solutions strategy. Previously, HP attempted to partner closely with BEA/WebLogic, but found itself in a position where it was in the passenger, not the driver seat, when it came to selling e-business solutions. HP realized that, to make its way to the table, it had to put skin the game.
For both HP and Bluestone, the acquisition didn’t come out of the blue. Both have been gradually ramping up their relationship since they announced an interface between Bluestone’s Total e-Business and HP’s ChangeEngine earlier this year. And the possibilities of leveraging Bluestone’s XML server capabilities could go far in helping HP flesh out its e-Services product framework.
Appserver Acquisitions: A Tale of Two Cities
The application server market provides good examples of what can go right—and wrong—from similar acquisitions. The track records of BEA and WebLogic; and Sun, Netscape, and NetDynamics, provide black and white examples.
Prior to the WebLogic acquisition, BEA itself was a collection of formerly independent and orphaned middleware tools. It was best known for rescuing Tuxedo, the leading third party UNIX TP Monitor, from Ray Noorda’s Novell, where it had languished as part of the company’s failed effort to become the center of the UNIX Universe. (At the time, Novell owned the remains of UNIX Systems Labs, the former AT&T rival to the Open Software Foundation.)
Back to BEA, the company struggled cobbling along a middleware strategy during the client/server era against the wrath of Oracle. BEA’s battle was uphill for other reasons as well. The TP monitor notion was both behind and ahead of its time, and it was an invisible technology that was marketed during a time when most enterprise IT organizations were preoccupied fighting the ERP wars.
For BEA, WebLogic itself came along at an especially auspicious time. With client-side Java reeling from performance and Microsoft-induced marketing problems, WebLogic was the first to get its arms around Enterprise Java Beans (EJBs). It would sell EJB appservers before EJBs were even a spec. WebLogic had compelling technology, but no feet on the street. BEA was looking for a hot product that would make the rest of its strategy make sense.
The result became, in effect, a reverse acquisition. BEA bought WebLogic, and although it kept its corporate name, the WebLogic product line surged front and center. BEA used its existing middleware products, such as Tuxedo, to lend more weight to WebLogic, and in so doing won the perception battle. WebLogic Enterprise may be a group of products, but in customer eyes, they often regard Tuxedo as WebLogic’s back end. Today, WebLogic is the leading J2EE application server in market share.
On the other end, the Sun acquisitions, which are now part of the iPlanet family, have suffered more lukewarm fates. Netscape originally claimed the mindshare for large-scale application servers largely thanks to its companion webserver and browser product lines. NetDynamics was known as one of the earliest application servers that used Java. However, because both products predated J2EE and EJBs, their internals weren’t Java-based. That in effect made them proprietary software products that just happened to use Java. The result was, when server-based Java standards emerged, neither could provide the degree of openness that was part of the Java promise.
When Sun bought NetDynamics, the company was one of the top players in the low-end appservers space. However, after it acquired NetDynamics, the company lost key personnel, and more importantly, market momentum, given Sun’s lack of experience in marketing software. By the time that Sun acquired Netscape’s application software business, spinning it off as iPlanet, it grew obvious that the NetDynamics’ product line’s days were numbered, given Sun’s strategy to tackle higher-end accounts. IPlanet’s execution, not its strategy, was flawed. It delayed in getting the message out to NetDynamics and Netscape customers, and in so doing, lost market share (in many cases, to BEA).
Since then, Sun has also folded Forte into iPlanet. Significantly, the management team that was responsible for Forte’s revival in fortunes did not survive the acquisition. Today, iPlanet’s software business continues to be run by the same management team responsible for the slow NetDynamics/Netscape consolidation. Significantly, the team lacks anyone experienced in the software business.
A major condition of acquisitions, especially those that look good on paper, is cultural fit. On one level, there shouldn’t be a problem: both companies have traditionally been very engineering-driven. But in the e-business market, that is not necessarily an advantage.
Another concern is the general question that arises anytime a large company swallows a smaller one: Will the buyer stifle the speed and agility of the buyee. The concern is especially acute for HP, which acted in characteristic, consensus driven fashion while courting Bluestone. Obviously, acquisitions aren’t casual matters, but once the deal is consummated, the newly joined parties must be braced to jump ahead at Internet speed.
From recent history, it appears obvious what must happen for HP’s Bluestone acquisition to pay off: Steal a page from BEA and make it a reverse acquisition.
On this front, the early news is good: Bluestone will become the headquarters of HP Software’s Middleware Division, with Bluestone president Kevin Kilroy taking over the unit. And, when it comes to e-business middleware, HP is almost a blank slate, with one obscure product and another that is still primarily high concept. Compared to other large recent acquisitions, Bluestone should have relatively few feet to step on. Lets hope it still has the feet to do the stepping. HP must hang onto Bluestone’s development organization, which knows a thing or two about change.
The Middleware Group should call the shots when it comes to e-business platform product direction, sales, and marketing, rather than waiting for consensus. There are obvious synergies with OpenView, which can be used to manage the back end systems that feed web commerce. And it shouldn’t shy away from rebranding: Who knows what ChangeEngine and HP Service Delivery really mean anyway? All e-business products should share a common branding—Bluestone Total e-Business and HP e-services and ChangeWhatever, alike. They need to mobilize all the feet on the street, from sales to service. The last part could be tricky, especially if HP’s crazy obsession to spend $20 billion on PwC diverts badly needed resources.
And, as long as we’re shaking things up, how about getting the web site up to date? Today, the web page containing the production description for HP Service Delivery lists Active Software’s ActiveWorks as the integration engine. For the record, WebMethods acquired Active Software back in May, and subsequently renamed the product WebMethods Enterprise.