An open alternative
Free OpenNMS has the right price but isn’t quite ready for larger, multisegment networks
WHEN EVERYTHING’S WORKING properly, a network manager is as busy as the Maytag repairman. Unfortunately, that nirvana has become harder to reach in recent years. Although current network equipment is more reliable than gear available five years or 10 years ago, the nature of network traffic is slowly shifting. As streaming media, especially video, becomes more prevalent, identifying and removing network bottlenecks will become more important than ever.
Traditionally, the NMS (network management system) is a big boy’s toy, found rarely in shops with less than a few hundred users — unless the smaller shop has an extraordinarily complex network. NMS vendors have tried Gillette’s old razor-blade-and-handle trick for years, by giving away core pieces of the software as “frameworks” into which customers can plug expensive, special-purpose modules, leading ultimately to a lucrative consulting contract to get the installation working properly.
Computer Associates Unicenter, Hewlett-Packard OpenView, and IBM Tivoli are examples of widely deployed NMSes with reputations as powerful, but hungry, beasts.
But there is an alternative path to paying for software and then paying to make it work. We took a look at a late beta release of the OpenNMS project’s eponymous open-source NMS that’s been under development since 1999; at press time, a Version 1.0 release was scheduled for April 8.
OpenNMS has an eye-catching price for its software: free. As do the established NMS players, OpenNMS intends to make money on consulting; unlike them, OpenNMS has a public, fixed-price listing for its services, starting with a simple one-day seminar and going up from there for more complex jobs. We found cost alone made OpenNMS 0.9.6 worthy of consideration, and the product’s performance on the Test Center network was icing on the cake.
As do the more established NMSes, the OpenNMS software relies on SNMP to provide communication between devices and the NMS software. The recent flap over security holes in SNMP hopefully reminded network managers that simplicity and security don’t mix, underscoring the need for a secure network management protocol. Nevertheless, SNMP is ubiquitous, and in the absence of anything better on the horizon, it’s the lingua franca of network management.
The core OpenNMS software runs on a Linux server; preferred distributions for the preview and the initial release include Red Hat 7.1 and 7.2 and Mandrake 7 and 8. We expect that future releases will add support for other Linux distributions; OpenNMS has been made to work on Mac OS X, and a port to Windows NT/2000/.Net Server is in the preschedule planning phases.
From a user’s perspective, the OpenNMS interface is browser-based and runs on Java for portability, if not always for speed; but the actual program code is rather light in footprint. As the number of monitored devices increases and data collects over time, so increases the need for high-end equipment in the role of NMS server.
Automatic or manual installations of the free OpenNMS software are relatively simple, assuming your IT staff has enough experience with Linux to get a server up and running. The only dependency that the installer couldn’t resolve on our freshly-loaded Mandrake Linux 8.1 server required us to manually install a Java virtual machine (VM) from IBM. Other than that, we were able to relax and let the script download, install, and configure supplementary services such as the Tomcat Java servlet package, the JBossMQ JMS (Java Message Service), and finally, a Postgres SQL database to manipulate the collected data before tackling OpenNMS itself. We wound up manually installing the three OpenNMS packages because of script failures, but that’s what we get for playing with prerelease code.
Once in place and configured with a couple of entries to an XML file, the OpenNMS poller quickly and accurately located network devices, and multiple interfaces were aggregated through DNS or SMB (Server Message Block) resolution or SNMP. Alerts via e-mail and pager were easy to configure. Of course, the true value of an NMS doesn’t show until it’s been in place long enough to draw a baseline for network traffic; the capability of documenting trends is critical for capacity planning. OpenNMS has spare but solid reporting facilities, including the option to automatically e-mail performance charts in Adobe PDF files to users and others concerned about the health of the corporate network.
Perhaps the most frustrating thing about OpenNMS is the inability to aggregate data from a number of OpenNMS consoles. This data consolidation is critical if OpenNMS is going to be useful for larger, multisegment networks, and it is expected to be included in OpenNMS 2.0 due out this fall.
This lack, but not much else, keeps us from recommending OpenNMS as an immediate replacement for the CA, Hewlett-Packard, and IBM offerings. But any shop with more than a couple hundred users should begin considering OpenNMS 1.0 as part of a long-term strategy for reducing the cost of network management.
