Running a network without monitoring it is like playing with fire: Sooner or later you’re bound to get burned.
But a number of IT departments feel they’re either too small or don’t want to pay for the best-known commercial products such as IBM’s Tivoli, HP OpenView, CA eHealth or Microsoft Operations Manager.
There is another option: Open source monitoring applications.
That’s what Xander Toushek turned to earlier this year after he was hired as a team leader in the systems department of Marketlink Solutions, a Toronto business consulting, Web design and IT hosting company.
The firm has 40 servers in two data centres with a storage area network, one here and another in New York City, 20 switches and 10 routers. It also manages two data centres for customers. But Toushek discovered it was using an “antiquated” monitoring application whose way of finding if a device was working was by pinging it. “We were always on the reactive end of failures and maintenance,” he said.
Toushek had been evaluating Zenoss Core, the free version of an open source IT monitoring system from Zenoss Inc. before joining Marketlink. Once there, he convinced the president to spend US$5,000 on the Zenos Appliance, a fully-configured Linux rack server running Zenoss Enterprise.
“It’s worked out extremely well,” he said. “Zenoss does full SNMP logging, sys logging, all your Windows-based events. With the executive view of the network you can actually see which devices are up, are down, which need attention for a system-level error. You know what the operation metrics are, what the thresholds are.”
While there are many open source network monitoring applications, IDC senior analyst Tracy Corbo says only a handful are truly open – that is distributed under a licence approved by the Open Source Initiative. They include offerings from Zenoss, Nagois, OpenNMS, Zabbix, Hyperic and GroundWork Open Source.
Some are entirely free, while others have versions with extra features than can only be purchased and include support. Depending on the features you need, they can be a legitimate alternative to packaged monitoring software, Corbo wrote in a recent report, and not just on price.
Open software has the ability to be customized, which can be appealing to those with special needs. Web sites of these vendors boast an impressive list of North American customers, ranging from state governments and a division of IBM to universities. Among them are some Canadian banks and cable companies. However, none were willing to be interviewed for this article so we can’t say exactly how the apps are being used. Still, for monitoring systems, open source tools can fill in gaps left by other solutions.
Here’s a snapshot of each:
• Nagios. A pure open source project. Can be downloaded separately or found in Linux distributions from Debian, Fedora, Suse Linux and OpenBSD Unix, as well as Groundwork Monitor (see below). Corbo says it’s targeted at network engineers, requiring some technical expertise before being deployed system. Still, it has an extensive library of monitor plug-ins.
• OpenNMS (www.opennms.org). Free. Like Nagios, according to Corbo, it needs some work before being integrated into your environment. Support and consulting services can be bought through the North Carolina-based OpenNMS Group . Tarus Balog, that company’s CEO, said the agentless OpenNMS can monitor “hundreds of thousands of devices.” It ships with some 40 monitors and includes escalation alerts and data collection. The goal of Open NMS 1.2 is to compete against HP Network Node Manager, he said, while the upcoming version 1.5 aims to compete against IBM’s NetCool.
• Zabbix. Free. Corbo notes its ease of installation makes it close to a turnkey software, which is its strength. Like Nagios, it can be found in several Linux distributions, including Debian and Ubuntu.
• Zenoss. CTO and co-founder Erik Dahl says the goal of the agentless Zenoss is to monitor the entire IT stack. While the Appliance can monitor up to 1,000 devices, the Enterprise software version can scale up to 20,000 devices. Dahl said the company has worked hard to make sure it plays well with Windows. Enterprise users also get ZenPacks, plug-ins that expand the system’s capabilities. Corbo notes that Zenoss Core is full-features, but Enterprise has an enhanced user interface, reporting capabilities and better multiserver management features.
• GroundWork . Going beyond performance monitoring, IDC calls it a framework for Windows, Unix, Linux, Netware and IBM iSeries system management. According to David Dennis, the San Franciso company’s senior director of product marketing, it integrates a number of open source tools, including Nagios. Comes in community, Professional (single location) and Enterprise (same features but supports thousands of devices) versions. Its default mode uses agentless monitoring, but for those who prefer there is an optional agent. Dennis said Monitor offers “on par” scalability with products from IBM and HP. GroundWork’s strength, says Corbo, is its ability to let users create a custom monitoring system.
• Hyperic . Stacey Schneider, the company’s senior marketing director, said Hyperic HQ (community) and Hyperic Enterprise versions manage over 70 application servers, databases and switches. The Enterprise version adds the ability to create templates for rules as well as high availability. Unlike others, which are priced by level of support, Hyperic is priced per CPU socket pair or network device. According to IDC, Hyperic’s strength is in its specialized agents for reporting detailed service information.
These applications have varying capabilities, IDC notes. For example, while Hyperic, GroundWorks and Zenoss can some ability to restart and reboot some servers and services, these are not built in to Nagois or OpenNMS. Zenoss and GroundWorks can show if there’s compliance with customer service level agreements, while OpenNMS, Zabbix and Hyperic keep historical availability reports.
Corbo adds another caution: Like all open source applications, just because a licence is free doesn’t make the cost of ownership cheaper. Support and training and deployment costs have to be factored in.
