Request for project specs to remove the big network lock

September 13, 2011 posted by Julio Merino

Dear users and developers,

The Board of Directors is interested in improving the performance of the networking subsystem of the NetBSD kernel on multiprocessor machines. To help people interested in working towards this goal, the board is willing to fund related projects.

The main idea behind any project that attempts to achieve the goal stated above is to get rid of the big kernel lock that surrounds the networking code. This would make the networking subsystem of the kernel more efficient on multiprocessor machines and thus allow the NetBSD kernel to remain competitive with other freely-available operating systems.

At this moment, we are asking for interested parties to submit project specifications including a description of the problem, a list of deliverables, a description of the tests and benchmarks that will be performed to evaluate the results of the projects and an estimation of the cost (monetary and timely -wise).

Please send any project specifications, questions or comments to board@ and core@ (both at

Note that this announcement has also been sent to tech-kern@ and tech-net@.

Update (Sep 14th): The deadline for project specs submission is October 31st. We reserve the right to extend this deadline if we are not satisfied with the proposals received until that date.

Thank you.

Julio Merino,
On behalf of the Board of Directors

[1 comment]



I have this sneaking suspicion that this last one is going to be one of the harder ones, on the basis that had it been easy, it would have been done before. But in cost:benefit terms, this *has* to be one of the biggest wins out there. 16 and 32 core platforms are now routine in rackspace, and no doubt 8 will move down into laptop land, where 4 is now not uncommon. If we were multithread across the board, a bunch of things might work better

Posted by George Michaelson on September 15, 2011 at 01:44 AM UTC #

Post a Comment:
Comments are closed for this entry.