* Readers (kind of) say no to agents
In the last newsletter, I shared readers’ comments on why they see software agents are more than just a necessary evil in their networks.
To them, the technology that gets distributed to servers and endpoints serves a critical purpose. But they also pointed out their hesitance with agent-less technologies unless they are only seeking a high-level view of performance or monitoring status. Without an agent, some say you can’t get the granular level of detail needed on each endpoint.
It’s a bit harder to pin down a clear stance in this week’s newsletter. Basically, as mentioned last time, agents perform some tasks that agent-less approaches simply cannot. So while people would prefer not to support a half dozen software agents on each endpoint, they feel in some cases the agent-less alternative simply won’t get the job done.
Here some readers shared why so many agents simply are not necessary.
One newsletter reader posed his own questions:
“Isn’t the idea of a ‘universal agent’ what SNMP is all about? Instead of reinventing the wheel, working to make it propriety and calling it a NEW feature, wouldn’t it make sense for vendors to expose their specific information in that vendor’s ‘enterprises’ MIB?”
A marketing executive at Indicative Software weighed in on the topic as well. The company uses agent and agent-less technologies in its service management software. While Indicative uses agents, this spokesman also extols the virtues of passive monitoring.
“Agents have their downside, but so does a ‘pure’ agent-less approach that depends upon newer health monitoring capabilities being built into hardware and applications. While we lean heavily on the idea that Indicative is ‘agent-less,’ we cache that in terms of using what we call data collection engines, or DCEs. They could be regarded as ‘proxy’ or ‘super’ agents (we’re keenly aware of a competitor marketing something similar as the latter), but at the end of the day they are designed to unencumber administrators from the mundane, ongoing burden of maintaining hundreds and thousands of agents for the sake of monitoring.
Indicative also has a pretty robust passive monitoring capability, and one that we of course believe exceeds the capabilities offered by competitors such as Coradiant and HP. What’s important to realize is that our real-user monitoring (RUM) module is seamlessly integrated with our core product architecture, which by design has extensibility to allow such seamless integration into a single product. It’s one of the banes of our existence: by virtue of being a single product that does what many point products are confined in doing, capabilities such as RUM get “lost” in the feature set. Nevertheless, we regard our appliance-based RUM module as an important complement to the active, or synthetic monitoring that provides the basis for proactive, problem-impact aversion our customers appreciate. The data collection engine paper I sent to you last week describes the all-but-agent-less approach that is the cornerstone of our synthetic monitoring capability.
