Look beyond your network

You'd think network management would get easier. Everything connected to enterprise networks is simpler than it used to be. Network standards make integration more straightforward. And enterprise systems management tools and other software programs put control of networks at administrators' fingertips. Right?Right. But still it isn't that simple. The number and types of devices connecting to enterprise networks have mushroomed. Wireless access points open networks up to a whole new class of security and reliability problems. And with more users connecting to networks from outside the firewall, and with more applications depending on reliable network communications, a network manager's life is anything but simpler.The main job of a network management tool is measuring network performance. Large-scale suites such as Hewlett-Packard Co.'s OpenView, and the command-center views in IBM Corp.'s Tivoli platform and Computer Associates International Inc.'s Unicenter, each provide some sort of network performance monitoring functionality, as well as device monitoring through the Simple Network Management Protocol (SNMP). They have for much of their existence.Those suites and platforms fall more into the category of systems management, of which network management is often a part, though it sometimes exists separately. The applications in this guide are among those that address a new set of needs.With the rise of Web applications serving users within an organization and over the Internet, there's a need for an integrated approach to managing networks, applications and systems from a single point of view.'Having one person look at applications and one [look] at the network didn't make sense,' says Frank Kettenstock, vice president of product marketing for network management for Concord Communications Inc. 'Customers are now looking for integrated solutions that can track both.'The first big effort to bring network management tools into the realm of unified management came in the form of enterprise systems management frameworks. The big ESM vendors'Tivoli, Computer Associates and others'sold many customers in the 1990s on the concept of a single framework to solve all their administrative ills.To a large degree, they delivered on this promise, but often only with the help of consultants, code and lots and lots of money. And the license maintenance costs of these systems drove many customers to scale back or abandon their use over the last five years, reviving the demand for 'best of breed' network management tools, and for a quick return on investment without major configuration headaches.One of the tools taking the out-of-the-box approach is Concord's eHealth series of software. The tools combine application, system and network management with the measuring of performance and availability of each component of distributed systems, and centralizes their control.The software combines the use of traps, which catch errors and system exception alarms sent by SNMP agents, and polling, which retrieves performance data from agent software, to create a unified picture of a network's health and performance.A problem with unified management systems is the potential for information overload. To be effective, network management tools have to reduce the amount of information that gets thrown at administrators and deliver a better quality of information instead.In the past, the failure of one network element could trigger hundreds of alarms, lighting a jumble of blinking red icons on the network manager's console. Only the manager's knowledge of the network could filter out all of the apparent alarms caused by the failure of, say, a single router port.To help network support teams quickly narrow down the scope of a problem, many tools can be configured with sets of rules that process network events based on a hierarchy'so that the primary problem surfaces first. This is a key feature to look for in new tools, because it allows junior IT staff members to attack problems effectively.Even though WAN monitoring has long been a part of overall network management tools, the increasing number of applications that depend on connectivity to systems outside the enterprise network'either over shared value-added networks or over the Internet'has increased the importance of tracking WAN connections.That kind of built-in intelligence is becoming a requirement for all types of network management systems. As more applications become dependent on distributed systems, and the potential points of failure multiply, keeping systems up at a five-nines level of reliability'99.999 percent'is becoming a default requirement. And doing it without adding to the already huge workload of administrators is just as important.With this sort of functionality high on customers' lists of demands, some systems management vendors are placing a new emphasis on what they call autonomic computing'building software and hardware systems that can perform routine tasks without network administrator intervention and take actions to recover from software or hardware failures instantaneously.Most autonomic computing efforts so far have focused more on systems management than network management. They have features such as automated user provisioning, which automatically assigns'and rescinds, when necessary'access privileges to users for applications and servers, and configures their policy settings for network security and logins.But the goal of building network management systems that can fix networks completely without human intervention'the vaunted self-healing networks promised by some vendors'are still a long way off.