 |
Heroix Longitude 3.0, starting at $299 per monitored
server There are a whole mess of products out there designed to help you monitor the various servers scattered around your LAN and WAN. They tend to fall into two categories: big industrial-strength products that require a substantial effort to set up, including installing agent software on every monitored machine, and low-end products that can only track a few statistics but that are inexpensive and offer immediate payback. It's been about a year since I looked at the first release of Heroix's Longitude product, which tries to combine the best features from the high and low end markets. Specifically, they've aimed for an easy to use and deploy product with fast ROI that still covers multiple platforms with comprehensive reporting. In that time they've pumped out two major upgrades, so I thought it was about time to run Longitude through its paces again on my own network. Ease of installation remains one of Longitude's strong suits. Pop in the CD, and fifteen minutes later you've got a running installation in your browser. The software is Java-based, and will install on either Windows or Linux. Out of the box, it has rule sets built in for monitoring Windows and Unix boxes, Apache, IIS, Exchange, SQL Server, Oracle, WebSphere, WebLogic, J2EE, and Exchange servers. Version 3 adds SQL Server 2005 to the list, as well as InsightManager, OpenManage, and Cisco routers. You can also define your own monitoring rules or keep an eye on event logs. Everything is built on top of industry standards as far as possible: Adding a new machine to monitoring is as simple as typing in the machine name or address and telling Longitude which applications are running on the box. Monitoring is all done through normal APIs such as WMI and JDBC, so there's nothing fancy to do to your application servers (though you may need to go back and install optional components - SQL Server, for instance, doesn't install its WMI provider in its default configuration, and you may need to turn on disk performance monitoring on downlevel Windows). Version 3 adds a nice new feature in the form of "correlated events": you can now take two or more events and create a synthetic event that only fires if the two of them occur together. So, for example, if you have a box where you only want to be alerted of high CPU load when a particular Active Directory condition exists, you can create an event representing just that combination. All of Longitude's reporting and monitoring dashboards are available via HTTP and HTTPS, so you can use it on your intranet or securely over the Internet. It also has role-based security and an e-mail alerting interface if you want to go that route. The reporting is exceptionally well-done; it's a whole series of graphs and meters with good use of color and everything is clickable so you can drilldown into detail. For example, you can pull up a report showing aggregate CPU usage across all your Windows boxes, and then drill down into individual hotspots and focus in on processes that are sucking down CPU cycles, with graphs showing change over time, in just a few mouse clicks. A lot of effort obviously went into this part of the product, and it works quite well. Version 3 also brings in basic "close the loop" capability with the ability to execute actions on monitored computers via command line execution as well as SNMP traps. There are a lot of other nice little changes here as well. SLA monitoring support, for example, has been in the product from the start, but now you have fine-grained control over SLA schedules to specify maintenance windows and other planned downtime. Another nice touch is the ease with which you can adjust rule thresholds to represent the actual activity in your own environment. When you go to manage the thresholds for a rule, Longitude will show you the historical data that it's collected for that rule, allowing you to decide what the normal range is on your network and set exceptional conditions with a good level of confidence. Although the basic Longitude architecture is agentless, it does make provision for an intermediate Statistics Server to collect data and forward it back to your main server. That way you can do store-and-forward consolidation from multiple branch offices or subnets if you need to cut down on the traffic generated by a pure agentless product. Pricing is on a sliding scale depending on which applications you're managing on each server, with a higher charge for some applications and volume discounts. One thing to watch out for is the minimum product requirements - you really do want to have the gigabyte of RAM in the monitoring server that it calls for. But given the appropriate horsepower, I found Longitude offered an easy-to-install and easy-to-use window into the operating health of the various servers on my network, and I was able to concentrate on how they were doing rather than fussing with the monitoring software.
Mike Gunderloy is the lead developer for Larkware and author of numerous books and articles on programming topics. Published May 23, 2006 |
