Date: Tue, 14 Jun 2016 09:20:17 +0200 From: Yuri Iozzelli <y.iozzelli@gmail.com> To: soc-status@freebsd.org Subject: High Performance P4 Software Switch - Week 3 Message-ID: <CAPSqqEkOWWMKHgOS53LZVrWqCuSLDmdAW8onwxMqte=6ij0ZuA@mail.gmail.com>
next in thread | raw e-mail | index | archive | help
Hi Project summary ======================== In this project I would like to implement a modified version of the reference p4 (http://p4.org) switch on FreeBSD, which uses netmap for faster packet I/O: the goal would be reaching the 1 Mpps order-of-magnitude speed (current reference implementation is limited to 150 Kpps for a simple l2 switch with 2 hosts). Week 2 ========================= I identified the main bottlenecks in the memory management, and wrote custom new and delete operator for the most allocated classes. I ran several performance tests and optimized the allocation strategy for the application. The new bottleneck seems to be the queuing mechanism, that is currently impemented with locks. I am investigating the use of a lockless queue. Links: Wiki page: https://wiki.freebsd.org/SummerOfCode2016/HighPerformanceP4SoftwareSwitch#preview Code (gsoc-* branches): https://github.com/zarghul/behavioral-model Regards, Yuri Iozzelli
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CAPSqqEkOWWMKHgOS53LZVrWqCuSLDmdAW8onwxMqte=6ij0ZuA>