Have You Been Lost?

I like to travel, which is a good thing since I just spent most of the last six weeks on the road. I was in Beverly Hills for a client, then England for a week with Arqiva. I got to visit with Jonathan Sartin (our ambassador to the UK and OGP member) and then headed off to Geneva. We have a large client there, and I got to spend a night in the French Alps as well as visiting the town of Annecy. After Geneva it was off to Naples to visit with Antonio Russo (our ambassador in Italy) who is working on some topology and map code. Unfortunately, I also got Lyme’s disease from a tick bite, and so I’ve been incredibly tired. This didn’t help me much since after the two weeks in Europe I was off to Texas, and I am now writing this from Arlington, VA (right outside of DC) and watching the boats on the Potomac.

I once had a friend named Lucy who worked at the local supermarket. She tried (somewhat in vain) to teach me Swahili, and I don’t remember much. I do remember that if I had been gone for awhile, she would greet me with a phrase that translated to “Have you been lost?”. I loved it. It implied that, of course, there is no way that I would have been rude enough to stay away so long unless something had happened: i.e. unless I had been lost.

So I wanted to apologize for not writing in such a long time. I eat, live and breathe OpenNMS, so I haven’t been away from the project, but I have been rather scarce on the lists.

There are lots of things going on. The week I spent in England was to demonstrate the high level of integration available with OpenNMS from within a framework. We coupled the alarms subsystem in the development branch with work Craig Gallen did with OpenOSS where he created an OSS/J interface to those events that could be accessed by any application that implemented the framework. The application that was demonstrated was called Statewise. This is a new application based on the Drools Rule Engine that did a really good job of taking a model of the network and isolating root cause, and via a JBoss application server it was able to respond to OpenNMS alarms.

I don’t believe Statewise will be released under an open-source license, however. I can hear the shouts from the peanut gallery from here, but please understand that the open source business model is not for everyone. I would rather work with a company that is providing a good application even though it is proprietary than to deal with those that are only using the term “open source” as a marketing tool. We were open source when it wasn’t hip, and it is my plan that OpenNMS will always be free as in freedom. The goal is to turn OpenNMS into the de facto management platform, and things like OpenOSS should go a long way to both convincing people to integrate with it as well as convincing companies to adopt it. I would love to see something like CiscoWorks run on top of OpenNMS, even if it is proprietary, and we can already integrate will tools like Netcool and InCharge easily.

The thing I like most about NGOSS is that, while it is from the telecom industry, it is a complete and well documented framework. I have longed for the luxury of designing such a framework and it was quite amazing to see it handed out on a platter. Like OpenNMS, it does have a steep learning curve, but the benefits are worth it.

Speaking of being a platform, we had a breakthrough today. Currently, we use one file for every value we collect in our data collection system. We used to use one file per device, but found that it was very hard to add new values without scrapping the old data. The downside is that the filesystem I/O required on large installs is tremendous. At New Edge Networks we are collection about 500,000 data points every five minutes, and this requires an NFS mount to a fast and large NetApp NAS.

Well, in the unstable branch, Matt and Dave changed it so that each group defined in datacollection-config.xml has its own file. For mib2-interfaces this means that the 11 values that are most commonly stored in 11 files can be stored in one. It takes just as long to write them as it did a single file in the past, so it represents an order of magnitude increase in our scale. Theoretically we could write 5 million values every five minutes.

Can I get a w00t ?

We have a few other surprises in the pipeline. Expect a 1.2.8 within 2 weeks and a 1.3.2 about the same time if not a little later. Open source is less about code and more about community, so thank you for making OpenNMS what it is. Check out our SF stats, 161 is the highest we’ve ever been I think: