Friday, November 22, 2013

vBSDCon Trip Report: John-Mark Gurney

The first trip report from vBSDCon is from John-Mark Gurney:

Thank you for the opportunity to go to vBSDcon.  vBSDcon was a great conference, and I was immediately and warmly welcomed by the host, Verisign, when I arrived.  They ran an exceptional conference and I hope this becomes an annual event.

One of the main reasons I attended the conference was to be able to meet FreeBSD developers I haven't seen for years and to meet new FreeBSD developers.  For the first time, I met Luigi Rizzo who I worked with over 15 years ago on FreeBSD's sound system and ISA UPnP code.  It was good to meet him and talk about some of his netmap work.

I also met some long time developers, Scott Long and George Neville-Neil, but I also met some that I haven't met before such as Adrian Chadd, David Chisnall (theraven), Ed Maste, Randall Stewart (rrs) and Baptiste Daroussin (bapt).  There are many more that I am forgetting, but that's simply because the conference was well attended.

I participated in the embedded BOF.  It was well attended and Adrian Chadd was the leader.  One of the issues that came up was the issue of how to make ARM and/or MIPS platforms Tier 1.  It was discussed how relevant the requirement that a new Tier 1 platform must be able to cross-build packages from an existing Tier 1 platform.  There was also discussions on which ARM board should be chosen was a reference platform.  The ones that were discussed as most viable were the BeagleBone Black, Raspberry Pi, a ChromeBook or a PandaBoard.

I attended the talk that Verisign gave on how they use FreeBSD to host the .net and .com domains.  It was impressive what they are able to do with netmap for UDP traffic, but they are also running issues w/ TCP connection performance on FreeBSD due to some lock scaling issues.

The hallway track is always interesting, and I was able to talk with David Chisnall about ways to possibly support changing functions at kernel load time to choose the best implementation.  One use of this would be to choose which AES implementation is best for the machine. If the machine has support for native AES instructions (AES-NI for example) it would use that, otherwise it could fall back to a software implementation.

Though unfortunately the hallway track made me miss Brad Davis's Speed Geeking talk on GELI, but it did allow me to spend time talking with Matt Olander, co-founder of iXsystems, Inc.  It was interesting to learn about the history of iX and find out about where they are interested in going with FreeBSD.  I may be able to help them with some work on FreeBSD in the future.

I attended the talk that Henning Brauer and Reyk Floeter gave on pf in OpenBSD.  The new features that pf has are very interesting and exciting. The features make possible reverse HTTP proxies and other packet steering techniques very easy to do.  I do wonder if using DXR for the table lookups that pf/ipfw do could be helpful, though the rule scanning overhead is still probably the largest overhead.

In summary, the conference was great and I was able to meet a large number of people and get some great ideas on how to help move FreeBSD forward.

Thank you again FreeBSD Foundation for providing me with the travel grant.

No comments:

Post a Comment