|
|||
|
|||
The Ones to Watch
Posted March 15, 2004
FTPOnline interviewed Rachel Chalmers, an analyst at The 451 Group, an analysis firm. Read about which application management vendors are making waves in this rapidly changing market. FTPOnline: How is the area of systems management changing, with many new startups entering the application management wing and older-school vendors such as Computer Associates, IBM, and HP moving toward them? Rachel Chalmers: One profound change has been the application of Bayesian mathematics to network and application performance problems. Thomas Bayes was an 18th century English mathematician and theologian who came up with a mathematical rule explaining how you should modify your beliefs about a system in light of new evidence about that system. It's a subtle and powerful notion now being applied to everything from clinical drug trials to spam filters—it underpins the popular and successful SpamAssassin. Bayesian inference gives you a way to think about information you don't know yet. Obviously it's a good way to tackle all kinds of problems in complex physical systems such as networks and applications. Companies such as Peakstone and Altaworks are exploring the application of Bayesian mathematics to multivariable performance problems with—maybe predictably—highly variable results. Larger Vendors Playing Catch-Up Rachel Chalmers: Larger vendors are catching up the way large vendors usually catch up—that is, by acquiring the best technologies they can afford. Mercury Interactive bought a company called Performant, which grew out of a University of Washington consulting gig to performance-tweak the terabyte database containing Boeing's 777 fly-through simulations. Quest Software bought a [Java 2 Platform, Enterprise Edition] J2EE expert company called Sitraka, and Veritas came pounding into the application market by buying the publicly traded Precise Software Solutions. FTPOnline: Which of the new startups have the most potential to shake up the status quo? Which are the most innovative? How do they differentiate themselves? Rachel Chalmers: In application performance and diagnostics, ProactiveNet Software and Wily Technology are fairly determined to go public if they possibly can; that would shake up some of the incumbents. Altaworks, Panacya, and ProactiveNet all have interesting takes on the dynamic thresholding or baselining problem, where their software learns in real time the application's normal operating parameters and then notifies managers when behavior falls outside that normal range. In fact, ProactiveNet has patents on various pieces of this approach, and believes Panacya might be in violation. The performance vendors also differentiate themselves on which technologies they can tackle. Many are J2EE deep dives, but others, like Heroix, play on the Windows side, or like Cyanea Systems, hook into CICS and IMS back ends. Still others, such as Covalent Technologies and Fidelia Technology, grew out of the Linux and Apache environments. Another dimension of the problem no one is yet tackling well is the human-computer interface issue: how to make sense of all that aggregated data. Various companies are trying to "percolate" or "bubble up" the most significant alerts and incidents, but it's a fundamentally intractable chore. A third dimension gaining importance daily is figuring out what you actually have installed—a much harder question to automate than it sounds. Companies such as Appilog, Cendura, Collation, Entuity, Intersperse, mValent, Relicore, Troux Technologies, and VIEO are doing interesting work here. Then there are the datacenter automation vendors such as BladeLogic, Centrata, Opsware, and Verbatix; database monitors such as BEZ Systems, Embarcadero Technologies, and Lecco Technology; network performance vendors such as ASG Software Solutions, Brix Networks, Compuware, Concord Communications, Empirix, Euclid, Lucent Technologies, Micromuse, NetScout Systems, NetQoS, Silas Technologies, Tonic Software, and Visual Networks; and service-level monitors such as Aprisma, InfoVista, Managed Objects, Packeteer, Sitara Networks, and System Management ARTS (SMARTS). The moral, such as it is, is that the performance or operations person's job is so incredibly hard that if you can automate even half of it, enterprises will beat a path to your door. No wonder so many entrepreneurs think management software could make them rich. They're right. An Advantage in No Hardware Agenda Rachel Chalmers: No. IBM, HP, and Sun are seen, not without justice, as selling management technology as a sweetener to move their overpriced hardware. To put it another way, the fads for reducing total cost of ownership by server consolidation or resource pooling or (insert buzzword here) are in fact different ways for the systems shops to pay penance to their enterprise customers who bought too much hardware during the boom, and who are now trying to make better use of it. This gives a BMC or a Veritas an advantage, in that they are seen as having no particular hardware agenda. CA is another story; it suffers from a long history of buying failing companies and milking their installed bases dry. CA's executive team says those days are over, but it's the rare CIO whose heart does not sink when a trusted vendor is acquired by CA. FTPOnline: What are some other large vendors to watch? Veritas or Microsoft? Rachel Chalmers: Veritas is a fascinating newcomer to this market, and all the incumbents are watching it with trepidation. Its Precise Software acquisition was shrewd. Microsoft is bound to come looking for application performance sooner or later, and the players on the .NET side of the house must be worried about that. I'm also interested in Mercury Interactive, which is turning Topaz into a force to be reckoned with. Top Goals for Operations Management Software Rachel Chalmers:Yes, yes, and yes. Which is the top priority depends on whom you talk to. Every test center and [network operations center] NOC has a subtly different problem causing the boss to tear his hair out. All these problems need to be solved, yesterday. FTPOnline: How do you decide on needs vs. complexity of product offerings? Rachel Chalmers: If only there were an easy way! The only way not to get burned is to begin the vendor selection process with a deep knowledge of the problem domain. First, understand the problem you need to solve. Then invite the likeliest suspects in for a bake-off. FTPOnline: Most of the problems with large, distributed applications occur not during development but in deployment. What can be done to address this problem? Rachel Chalmers: Unfortunately, there's no silver bullet. The only way forward is the hard way: better development processes, better test and quality assurance thanks to more complete and accurate staging environments, and better post-deployment monitoring and test tools. FTPOnline: While client/server tools are mature, IT managers often complain that managing distributed objects in production environments is difficult, and the tools aren't yet up to it. What needs to be done, and whose tools show promise? Rachel Chalmers: There's a difficult engineering tradeoff to be made between heavyweight agents that pull a lot of useful information out of objects, while exacting a substantial performance hit, and so-called agentless approaches, which don't get as much information but don't slow things down as badly, either. Again, everyone solves this problem in a slightly different way. From the buyer's point of view, it's horses for courses: deep dives like Wily in quality assurance and testing, perhaps, with a lighter-weight approach post-deployment. Many people criticize Wily for being an application-server deep dive, but Wily has a great comeback: Those who criticize the deep dive, can't deep dive. About Rachel Chalmers |