NYCBSDCon Summary Part II

Will Backman continues his journal from NYCBSDCon 2008 (previous entry)...

Day two of NYCBSDCon was another early start, although a little more coffee was needed than Saturday. The fantastic weather continued to hold, paving the way for a full day of interesting talks. Attendance was light in the morning, but there did seem to be more people manning the vendor booths. I made my way to the lecture hall after grabbing some of the catered breakfast.

After the kick-off, Theo Schlossnagle demonstrated Reconnoiter. The project was born from a frustration with the lack of scalability of the more common open source network monitoring projects. The project isn't yet complete, but showed a lot of promise. The web interface was amazing, and the option of a IOS-like command line interface was a nice feature for network administrators. I was glad to see that a lot of work went into building everything from non-proprietary components, and I'm looking forward to trying this out once it (hopefully) hits the ports tree.

The next talk was by Pawel Jakub Dawidek, where he introduced the ZFS file system. The first half was a description of some of the most powerful features of ZFS, which were presented through an animated presentation using cartoon characters drawn by Pawel. This was then followed by a more technical description of how ZFS was ported to FreeBSD and how it fits into the GEOM framework. There was a lot of discussion afterward, including questions about compatibility with other ZFS implementations, comparisons with HAMMER, and system requirements. When asked about production readiness, Pawel stated that he would like to see more developers working on the port of ZFS before he would be willing to take it out of beta.

Next was a presentation on OpenBSD's implementation of Position Independent Executables (PIE) by Kurt Miller (kurt@). He talked about how programs and data are traditionally stored in the stack, and how PIE can help prevent certain attacks by shifting locations in an unpredictable manner. Work is still ongoing, and performance is a concern, but the hope is that this method can be made a default option in OpenBSD in the future. During the discussion, issues with specific programs such as Emacs or Apache were mentioned.

During lunch, it was time for some live broadcasting again. I spoke with a few vendors, picked up some more free stuff, and purchased my OpenBSD 4.4 CD set. There was also a PC-BSD BoF, which was a nice chance for people to talk directly with Kris Moore about the project.

Following lunch was a presentation by Mike Silbersack on Detecting TCP regressions with tcpdiff. This tool, along with a fairly complex virtual machine setup, was used to find bugs in commits to the FreeBSD TCP stack. It was clear that this was a very useful project, as it detected some bugs that had been in the stack for many months. One developer in the audience suggested that this be brought into the tree so it could be used more often.

The last talk was an energetic description of Network Refactoring by Michael Lucas. He described his approach to evaluating and prioritizing network changes, working with users and management, and staying sane in the process. His description of the network that he inherited was astonishing, yet in many ways familiar to many in the audience, and his recipe for successful change was very helpful.

The closing remarks were a chance for the conference organizers to say thank you to everyone who made the event possible, and also a time for the much anticipated prizes such as t-shirts and some neat single board computers from PC Engines. For the more studious among us, BSD Certification Exams were offered at the end of the day.

Once again, NYCBSDCon was a fun and interesting conference. It was nice to see so many people from the BSD community come together and exchange ideas, and I can't wait to see what new ideas are spawned from these interactions.