When Jerry Clotfelter grew weary of looking after his crop of network servers, he looked for some software to automate the process. Unfortunately, he found nothing; it was back to the DIY method for him.
"I wanted to stop baby-sitting all six of my servers by hand," says Clotfelter of the process that kept him on call 24 hours a day.
The result of Clotfelter's work has not only brought him uninterrupted sleep, but also spawned a new Web-based remote monitoring service called RedOwl. RedOwl uses software Clotfelter wrote that, when placed on a server, automates the process of monitoring his network 24 hours a day.
Should a server ever go down, it's not necessarily Clotfelter who gets the call. Instead, the software he wrote sends out an email message to network administrators designated for such alerts. For US$30 per month, Clotfelter can set his server to watch someone else's network servers remotely. Now if only the customers will come.
RedOwl fills a big void in the field of remote network monitoring. Software that performs this task isn't likely to come in a shrink-wrapped box; rather, it is written on an ad hoc basis by network managers right when they realize they need it. But because these scripts are often left behind when an administrator leaves a company, the software and the methods of monitoring are not widely circulated. RedOwl is an attempt to help network managers get past having to reinvent the wheel each time they realize a need for remote network monitoring.
"I believe in the 'anything for a friend or a fee' philosophy," explains Clotfelter. "We built the service for our use, but if it is useful for others, I'm happy to sell it to them."
Network monitoring software works by allowing the server - in this case, Clotfelter's - to access a remote network as though it were any user of that network. For instance, to monitor a Web server, the RedOwl software will periodically attempt to connect to the Web server port and download a page just as any user would. The software will silently continue to do this until the Web server doesn't send the page, at which point the monitoring software should alert an administrator.
RedOwl is unique in that it performs this function from a remote location, a much-needed service because it is sometimes difficult to detect that a server is unreachable when monitored internally. Generally, when network administrators buy or write a piece of network-management software and run it on the local network, they can monitor the availability of their server from within their own network. If this network becomes disconnected from the Internet, the server may still function correctly but be inaccessible to the outside world. Because the monitoring software is run locally, it will not detect this condition and won't send out the appropriate alerts.
In addition, some corporate networks use the Internet as a transport mechanism for email. In the case of these companies, the alerts are supposed to travel from the local server, across the Internet, and on to the network administrator. With local monitoring, if the corporate network becomes separated from the global Internet, this process can't happen. But with the monitoring software on a server outside the corporate network, the network failure will be detected. An alert could then be sent from the monitoring site across the Internet to the administrator - all because the monitoring server is remote and therefore still online.
For the same reason, there is also a downside to many remote monitoring schemes. The monitoring scheme is only as good as the network connection between a remote monitoring site and the server being monitored. When a remote machine monitors a server, it is also monitoring the entire network between the monitored site and the monitoring site. This means that alerts are sent even when the problem is not on the monitored site's network. Most administrators would consider these alerts to be false alarms, and will tolerate only a very few of them.
RedOwl solves this problem by monitoring from three different locations, served by three different bandwidth providers. Alerts will only be sent if an error condition is verified from all monitoring points.
And that makes for a more reliable Web.
