"What we're trying to do," said Avocent's Steve Rokov when I spoke with him last week, "is remind Windows IT Pro readers that, sure, there's a whole bunch of existing agent-based technology that you should be using in your network-management duties, but there's also a next-generation wave of things going on underneath the covers that you really need take notice of."

You're probably constantly asked to do more and more with fewer and fewer resources, and at the same time, you need to ensure the highest availability of your network. More than ever, you need to have the right tools for not only monitoring and managing the network and its servers but also for quickly recovering the network when it goes down. The best way to have a sure hand over network management and uptime is to have the right mix of agentless and agent-based server-management technologies. Having agentless alongside traditional agent-based software management suites helps you cover the gaps traditional management software might leave behind.

There are many industry standards (Intelligent Platform Management Interface—IPMI, Systems Management Architecture for Server Hardware—SMASH, WS-Management) that combine to make up agentless, and its key benefit is that because it runs on a dedicated service processor, it runs as an independent subsystem. That opens up all kinds of possibilities, the biggest being that it fills operational gaps left from traditional agents that rely on the OS being up and running.

"Agentless is an interesting term in the industry," said Rokov. "In the past, many people have used the term to describe something that simply comes with the system. They might say, 'Isn't that agentless? I haven't procured it or configured it.' But many such tools are rudimentary, with no decision-making capability such as logging and setting thresholds or policies."

Agent-based and agentless are similar from an implementation view. Many agentless features reside in the hardware, complementing what goes on in the OS and applications. "In a typical boot-up scenario," said Rokov, "you have boot, initialization, OS load, and then you're up and running. At that point, much of the agentry of management software installs and admittedly provides a lot of value. However, you've got other situations where something might have happened to services, the OS, the application, and suddenly you're in a state where everything is hosed. Also, during boot and OS load, you've got no agent-based capability running yet, so you're kind of blind at that point. There's still a lot of time left for something bad to happen before the software agents begin to load. What do you do?"

Agentless fills those gaps.

No matter which state or condition the system is in (pre-OS, boot, distressed state), agentless is going to provide some additional value. Agentless doesn't require additional software and doesn't need to be configured as extra-complex software that sits in the OS, so it's a simpler method to add value to a server. Best of all, agentless tools are typically pre-integrated by the vendor, so they're free--"as much as free can be in this world," said Rokov. "It's like a free insurance policy. Isn't that marvelous?"

How about an example? Rokov cited a university with a supercomputer cluster running on UNIX and 500 mostly Intel-based servers. "It's an academic research community relying on the cluster to be up for research projects, so uptime is critical. Rather than paying for additional software-agent componentry, the IT administrator simply switched on the agentless capability to provide basic monitoring. Agent-based would probably have provided some of the capability he was looking for, but he really needed only low-level physical monitoring and control, and agentless was able to provide that. The great news was that by doing so, he saved the cost of a whole IT person in that process."

Avocent is a member of IPMI 1.5, a forum of more than 180 vendors supporting the standardization of agentless technology. "What can IPMI do?" asked Rokov. "If you have IPMI on your system, you have power control, you can look at hardware inventory, you can store critical events in the system for real-time use or as a way of seeing what's going on in the network, you can poll chips for temperature readings, and you can set thresholds and alerts. Rack hotspots are a real concern. Now you can pinpoint them earlier. You can predict problems before they happen."

"This stuff is just sitting under the covers, waiting for you," Rokov said. "Go ahead and use it!"

Avocent's MergePoint appliance provides security and management capabilities. For more information, visit the Avocent Web site.