The Interactive Network Design Manual The Systems Management Dimension by Bruce Boardman Architecture System management suites enhance their functionality through consolidation and integration. The goal is flexible control in either a centralized or distributed environment. The suites and add-on products have as a designed goal the physical and/or virtual centralization of control. The idea is to provide control no matter the environment-mainframe-centric, LAN-based or any combination of the two. This pretty much throws away the cookie-cutter categorization of management suite functionality, and the judgment as to the best architectural approach. Even with this value analysis rooted in the legacy and direction of a particular shop, understanding the basic services that are or should be part of a system management architecture is a good first step in building the right management infrastructure. Server/Agent All management products use servers and agents running on management targets to apply the various management disciplines. In the most basic form the server portion is made up of a console, database, discovery engine, polling engine, and event and correlation engine. The agent portion is simply a started task on remotely managed targets that respond back to the server with management application data and events. This rather simplistic approach, though working well in a centralized homogeneous environment, is not well-suited to distributed heterogeneous environments. A better approach is one that supports multiple standardized databases, distributed management applications through management application modularity, heterogeneous console operating environments and intelligent distributed agents. In the past a relational database management system (RDBMS) served one or two management functions with little flexibility as to the database type. Now not only is standardized database support to be expected, but so is the correlation between management application's data stores. One approach is to have a single centralized database scheme that applies to all management applications. This approach, however, lacks flexibility for the addition of new management applications and database types, requiring an incredible amount of data migration by the system management vendors. A second approach is to build on top of the separate databases a reference database that maps all of the management application specific databases. This takes two forms, either object orientation or repository. In the first case, object orientation, though flexible and extensible, is not yet ubiquitous even though it might be standardized. This is one of the big questions that will be answered over the next 12 to 18 months. In the case of the latter, a database repository is used to map and replicate the various data bases. This approach does not have as much potential for open multivendor development, but it a nswers the standardization question posed by the object oriented approach-by ignoring it. Distributing console access over multiple flavors of Unix is usually a matter of X Windows, and is a very basic and clunky approach. The control of services on NT servers from Unix consoles and vice versa is more flexible. However, few product have enough flexibility to support all of the permutations in a distributed environment that contains everything from NetWare to mainframe. Browser-enabled management consoles provide the answer for managing everything everywhere, and are an option to the better management platforms. Also part of distributing the management console is the dimension of the interface, character, 2-D, or even 3-D. Don't underestimate the advantage of being able to make a terminal link to a management server when dialing in from a remote location. Character interfaces may not be intuitive, but they are very efficient in the right hands. 3-D interfaces, on the other hand, rely on specialized video accelerator hardware, and for the moment appear to be more useful in making marketing pitches than actually solving problems. Their promise, however, is to flatten the learning curve, through increased intuitive navigation, thus making more problems solvable outside of centralized technical expertise. Agents are beginning to acquire more server functionality, being able to poll subagents and correlate local and subagent events. This can mean the decentralization of control or just the decentralization of server functionality. In the first case centralized access policy can delineate part or all of a management discipline to distributed local control, with updates forwarded to a centralized database. In the second case all control can be maintained centrally with automated correlation and actions performed by the distributed agent. The approach taken will depend on the management application and what fits into your shops' business model. With all of this distributed intelligence floati ng around, just managing which agent is where and what access policies are applied to it can be a daunting task. It is therefore important that asset management software not only recognize the existence of distributed management agents, but that a clear and easy way to identify on an agent-by-agent basis, (as well as on a policy basis) what functionality is configured for each distributed agent/management node. Get the MF in Here! The term "enterprise" means many different things depending upon who is doing the talking. For many shops interested in and able to justify the kind of effort in implementing large scale systems management, a mainframe (MF), is part of the picture. As we mentioned at the outset, mainframes have established support and control infrastructures, with well-defined disciplines that often don't easily translate to the client/server model. For this reason the most common approach taken for systems management integration of mainframes and client/server is to open a line of communications that supports event-forwarding and correlation between the two environments. This segregation maintains the disciplines to manage each, but it does not fulfill the goal of managing everything from everywhere. A second approach places an intelligent console over both mainframe and client/server management disciplines. It combines event correlation, command, security, storage, problem management, job control, performance, network management and topology into a single view. Updated December 17, 1996 Print This Page E-mail this URL