Arrow of time
Arrow of time

BSDCan 2012 - DevSummit

Share Tweet Share

Another year - another BSDCan! It's very nice and even comforting to see such a large number of familiar faces again …

Another year - another BSDCan! It's very nice and even comforting to see such a large number of familiar faces again, and even more as the ranks are filled in by fresh new developers. The conference and the Developers' Summit before it promise a great program and a great time for the BSDers.

FreeBSD as a project is blessed by a comparatively low number of kooks and freaks - which is probably a natural consequence of its long history. All the people I see here at the DevSummit are sober and reasonable, putting technical excellence in front of all other consideration. It is very close to an ideal meritocracy.

Not that there aren't cabals strewn around, because people will always be people, but they are of the "birds of feather" kind, held together by a common interest or few, and the barrier to entry is very low - expressed as "show us the code." As an engineer, I'm very happy about it.

This year's summit is organized during three days, overlapping a bit with the conference itself, and the days are packed with interesting topics. The events themselves usually follow a simple pattern: a group of interested developers meet in a lecture room and talk about specific topics, usually unde a guidance of few group leads for the topics themselves. Through the years it looks to me that these events have grown more and more formal - when I joined there were fewer developers and the topics themselves were esoteric enough to keep most of the people out, but now, they are more like invited, keynote lectures about the problems in certain topics which are presented by a few extremely knowledgable people to a larger audience. This way is more benefitial to spreading the knowledge around (and this is precisely why I like them) but there's a certain note of group work lost, delegated instead to the nightly "Developer launges", informal gatherings where real work gets done. Either way is fine. If the organizers wanted to keep tighter groups, they would have easily narrowed down the scopes of the topics, i.e. to "Let's break ifnet" instead of "Let's list all of the problems with the lower parts of the network stack".

I was at the offload engines / network stack track today, which have both completely filled the allocated meeting room, showing just how much interest there is in advancements in these areas.

The offload engines track was mostly about presenting the work of Jim Harris, an Intel engineer who is bringing in the support for Sandy Bridge's many special offload capabilities, as a basis for a future, formalized framework which would be multi-platform.

The network stack track basically announced a Grand Work on the modernization of the lower layers of the network stack - those dealing with physical devices - in order to make better use of queues (physical and virtual), TOE and modern protocols. I'm of the opinion that the majority of users would benefit more from reworking the upper part (TCP/IP, etc.) in order to scale on multi-CPU setups, leaving the lower layers a bit crippled in the edge cases for now but apparently this topic Will Be Discussed At An Appropriate Time. Of course, in the long term they are both going to be done.

Like always, these will be exciting days for FreeBSD!

I'll also be at the conference, giving a talk about Bullet Cache.

comments powered by Disqus