|
|
               
|
Masters of management
By Jim Duffy It's a faint sound off in the distance, but that buzzing is unmistakable. Suddenly - and sometimes with little warning - it's upon you. It's coming from those clever, handsomely compensated marketeers with all the flip charts, flashy brochures, slide shows and videos they think are needed to get you revved up enough to jump on the bandwagon for the next-generation network management solution. That's right. They want to sell you a Web-based tool that runs on top of a case-based reasoning platform. It's been integrated with all the best-of-breed Java- and object based applications that third parties can provide. The goal is to assemble an out-of-the-box solution that will enable you to truly manage the type of IT service levels those cable testers and SNMP device monitors only claim to control. Oh, yeah, it also conforms to all the standards du jour.
After the presentation, all you can say is, "Huh?''Exactly. Welcome to today's world of enterprise management where, unfortunately, the myriad tools, techniques and technologies grab more headlines than the actual problems you're trying to solve. Trouble is, with new management methods emerging all the time, keeping track of what's real and what's vapor can be more effort than sticking with the work-arounds that tend to get you through the day. So next time the glitz crew comes your way, be armed with a buzz zapper and use it whenever you hear the terms "service-level management," "object-oriented frameworks," "integrated management platforms" and "artificial intelligence (AI)". These terms are too ambiguous or don't relate to anything that will benefit you. On the other hand, perk up when the conversation rolls around to Web-based tools that can ease management chores, policy-based management, rules-based inference engines and case-based reasoning. If there's any management buzzwords worth paying attention to, these are the ones.
OverspokenBe especially quick on the buzz zapper whenever you hear the term service-level agreement (SLA) management. All of a sudden, companies that once sold products to simply monitor network device status, system CPU usage or disk space capacity have gizmos that guarantee you will receive expected levels of network throughput and application performance. However, if one of these SLA gizmos tells you you're not getting the performance you expected, don't be surprised to learn the company that sold it to you expects you to buy one of its device monitors so you can figure out where the problem is and what to do about it. However, that's not quite the true definition of SLA management. Truth be known, SLA management is not new. Crafty marketeers looking to extend the life of device monitors borrowed the term from the telecommunications world, where it's been used for close to a decade in a real sense. SLA management relates to your ability to monitor whether promised service performance guarantees and committed information rates (CIR) are actually being delivered. Even in the telecommunications world, it's still up to you to keep the carrier honest by tracking whether guarantees and CIRs are delivered and to fight for a rebate when they're not. Only recently have some carriers tried to make this task easier by enabling you to tap into their management systems to get the real-time service performance information you need to make a judgment. Give the buzz zapper some overtime to keep someone from blathering on about an object-oriented management platform, too. There was a day when all management platform vendors talked about making their products object-oriented. Today, few of them are. But does that really matter to you? Probably not, especially when you consider that an object-oriented management platform or framework really doesn't help you do your job faster. Instead, object orientation's benefit goes to
developers who can tap its code-reuse abilities to write new
applications faster. Objects won't really add new functionality to
the platform itself. They just will give you new applications
faster. Sure, some of those applications may make your life a
little easier because they can use management objects, but the
technology itself provides more of a benefit to the developer. There is likewise no need to get snowed by the term "integrated management product'' because management products are anything but integrated out of the box. Users say that deploying a vendor's management framework with applications from third-party partners still requires a significant systems integration effort. This is the case even though these partners' products are supposed to be "integrated'' with the platform from the outset. The result is frustration. "We can get [the functionality we need] from third parties, but I don't want to be an integrator,'' says Dennis Fishback, manager of data processing for Virginia Power Co., in Richmond, Va. Some users see relief in the recent initiatives announced by market leaders Computer Asso-ciates International, Inc. (CA) and Hewlett-Packard Co. to bundle stripped-down versions of their management platforms with systems and servers from multiple hardware vendors. "Having vendors do that is great,'' says William Oris, a vice president at investment giant JP Morgan & Co., Inc. "It lessens the amount of integration work we have to do.'' Yet the "platform'' model of enterprise management itself is quickly losing steam. In addition to the lack of integration, monolithic platforms require too much development time on the part of third-party application developers who have to pick and choose which platforms to support. If they want their applications to run on all the management platforms out there - and there are quite a few - developers have to write a different version of the same application for each one. This is an expensive proposition for application developers, and they usually pass that expense on to you.
Good termsEnter Web-based management. The World Wide Web's much-touted promise to make application portability a reality is genuine because Web browsers and Web-based applications, especially those written in the Java language, can run on any platform. "The Web paradigm is certainly the most scalable, the most ubiquitous,'' says Stephen DeWitt, vice president and general manager of network management at Cisco Systems, Inc. "It gets us out of the age-old, nauseating problem of porting to every platform under the sun.'' Consultants agree. "One of my beliefs is things like the Web are going to be the glue, the integration point, rather than the management platforms,'' says John McConnell, president of McConnell Consulting, Inc., in Boulder, Colo. Because of its familiarity, portability and user-friendliness, the Web browser is becoming the universal user interface for management consoles. With a browser and the Web, management data can be sent to anyone who needs it. This data is mostly in the form of HTML-based hot-linked performance reports. But more and more, companies are giving their network managers an interactive, real-time, Java-based Web interface that provides a greater graphical depiction of the managed environment that can be updated in real time. "We're pleased with a Web interface,'' says HP OpenView user Sandra Potter, network specialist at Air Products and Chemicals, Inc., of Allentown, Pa. "That helps us a lot. But the second phase is the Java stuff. We're really interested in third parties getting on board [with that].'' She's not alone. "Java will allow us to get [management] information to more people,'' says Paul Edmunds, senior network analyst at Duke Power Co., in Charlotte, N.C. Java, he says, can lessen the performance hit on management platforms that currently use X Windows. There are even two standards being proposed for Web-based management: the Web-based Enterprise Management (WBEM) initiative, backed by BMC Software, Inc., Cisco, Compaq Computer Corp., Intel Corp. and Microsoft Corp., but championed mostly by Microsoft; and the Java Manage-ment Applications Programming Interface (JMAPI), the brainchild of Sun Microsystems, Inc. The WBEM concept defines three areas for standardization. One is a HyperMedia Management Schema (HMMS), an extensible data model for representing managed objects. Since WBEM's introduction last year, HMMS has metamorphosed into the Common Information Model (CIM) defined by the Desktop Management Task Force (DMTF). Another WBEM component is the HyperMedia Management Protocol (HMMP), an HTTP-based protocol for communicating between management services, applications and agents. The last piece is the HyperMedia Object Manager (HMOM), a C++ object broker that will pull together management data on behalf of management applications. HMOM is based on Microsoft's OLE technology. While the DMTF is defining, maintaining and updating CIM, HMMP is being debated within the Internet Engineering Task Force, and Microsoft has placed a reference implementation of HMOM in the public domain. To date, there has been little, if any, activity around HMMP. Vendors were expected to release WBEM-compliant products by now. But don't hold your breath. WBEM is seen by many industry insiders as a ploy to get you to buy into Microsoft products. As such, vendors will likely only pay lip service to WBEM while pumping out Web- and Java-based tools independent of it. Speaking of Java, Sun is one of the vendors conspicuously absent from the WBEM bandwagon. That's because Sun and Microsoft are at odds over how Web-based management should evolve. JMAPI is part of SunSoft, Inc.'s
Java-based Solstice Workshop, a programming environment for
developing Web-based network and systems management software. A number of third-party vendors have endorsed JMAPI, including management rivals CA and IBM's subsidiary Tivoli Systems, Inc., because of Java's "write-once, run-anywhere" properties. Because Java is real, you can expect vendors to fulfill their promises to deliver JMAPI-based software later this year and early next year. In fact, Tivoli is committing to ship its Java-based Tivoli Management Environment (TME) framework in mid-1998, according to CEO Frank Moss. Tivoli was one of the early adopters and proponents of object-based enterprise management and even contributed object technology to the Open Software Foundation's ill-fated Distributed Management Environment. But Tivoli now acknowledges that the Web is the next wave. Web technologies aren't the only real hot button in enterprise management. Policy-based management also is getting a lot of attention, and rightly so. The ability to establish policies for how your network will be managed - defining what application will be launched based on certain events or establishing network access rights based on user profiles - will go far in automating management or making networks more self-manageable. The major internetwork vendors - Bay Networks, Inc., Cabletron Systems, Inc., Cisco and 3Com Corp. - all support policy-based management in one form or another. Also, the major enterprise management framework vendors - CA, HP and Tivoli - support policies for monitoring network and systems usage and health. Policy-based management could have been augmented by AI, that "next great thing'' of the mid-1980s. Trouble was, users were so squeamish about letting computers decide how to manage their computers, AI never took off. So AI in network management these days is being called rules-based intelligence and case-based reasoning, which really isn't AI at all. Rules-based tools foster automated management by initiating alarms or remedial action via predefined rules programmed into a management station by - you guessed it - people, whether you or a vendor. Case-based reasoning tools are essentially a repository, or library, of problem-solving expertise culled from your staff or the tool vendor's staff and programmed into an application. Case-based reasoning enables you to view present cases or develop a case and decide how to solve a problem based on information stored in the repository. The computers used with rules-based and case-based reasoning tools are not thinking for themselves as they would in true AI. Rather, they preprocess data so you can solve problems more quickly or automate responses to conditions you've programmed in. There are a couple of variations of rules-based intelligence in network and systems management: knowledge modules and inference engines. Both require that rules for inferring causal relationships be predefined by vendors or users. Knowledge modules are behind BMC Soft-ware's Patrol application management suite. These software modules are loadable libraries of expertise containing rules for how Patrol will manage an operating system, database or application. BMC predefines knowledge modules for specific environments, such as Lotus Notes, SAP R/3 or Oracle database management systems. Inference engines are exactly that: They provide a way for a management system to infer the likely cause or remedy of a network or system fault from the events that result from the fault. This ability to infer is preprogrammed into the engine based on the past experience and knowledge of a network manager. If a new method is used to remedy a fault, the manager updates the inference engine with this new cause-and-effect scenario. Seagate Software, Inc.'s NerveCenter event correlator is a good example of a rules-based inference engine. NerveCenter uses color codes to display varying user-defined degrees of event severity on the management console screen. It also can initiate user-defined responses to error conditions, including sending an e-mail, dialing a pager number, automatically starting corrective actions by launching applications and displaying instructive text for users. Case-based reasoning works under a different model. With case-based reasoning, the network manager can peruse present cases to determine if a similar fault has occurred and been previously fixed. Or, if a new method is needed to fix a problem, the manager can develop a new case and enter new problem-resolution information into the case-based reasoning repository. Case-based reasoning is available as an option on Cable-tron's Spectrum management platform. Cabletron's case-based reasoning tool also can interface with trouble-ticketing, help desk and call-tracking systems so users of those systems can tap the problem-resolution repository for more rapid troubleshooting and escalation. So what's the benefit, if any, of rules- vs. case-based reasoning? Case-based reasoning has the ability to retain knowledge that is otherwise lost because of personnel turnover and attrition. Case-based reasoning also is adaptable to a particular network operation because managers can update the case-based reasoning database with new techniques. This makes case-based reasoning more advanced than rules-based reasoning. Because rules-based systems rely on predefined rules or conditions to deal with unforeseen situations, they lack the ability to adapt existing knowledge to a novel situation, or to learn from experience, which is what a true AI-based tool is supposed to do. And learning from experience may be the only way you can differentiate hype from help in enterprise management. Maybe those that learn from experience can develop that Web-based, case-based reasoning platform that's integrated with best-of-breed, Java- and object-based third-party applications out-of-the-box. And maybe this gizmo can truly manage IT service levels that the cable testers and SNMP device monitors only claim to control.
How to Advertise | Copyright
Home |
NetFlash |
This Week |
Industry/Stocks
|
