3 ms·
Scott Long has summarized it a few times, eg, from http://people.freebsd.org/~scottl/Netflix-BSDCan-20130515.pdf http://people.freebsd.org/~scottl/Netflix-BSDCa
by brendangregg 12y ago
Scott Long has summarized it a few times, eg, from http://people.freebsd.org/~scottl/Netflix-BSDCan-20130515.pdf http://people.freebsd.org/~scottl/Netflix-BSDCan-20130515.pd...:
• Availability of expertise, outstanding community
• Works well, good vendor support
• No GPL
• Features used:
– SUJ
– gmirror – boot drive only
– AIO
– Dtrace, HWPMC
– TCP Stack, modular CC
It's missing "Integrates tightly with nginx: Async I/O and sendfile()"
which Scott included in http://youtu.be/FL5U4wr86L4?t=18m30s http://youtu.be/FL5U4wr86L4?t=18m30s .
In the video he elaborates a bit more about the GPL...
edit: formatting
- cbsmith 12y agoWouldn't you say that Linux integrates just as nicely with nginx's async I/O and sendfile support?
- m0th87 12y agoI thought *BSDs used kqueue, which allegedly had worse performance at high concurrency vs Linux's epoll? EDIT: This seems to suggest kqueue may have better performance due to less syscall overhead: http://www.eecs.berkeley.edu/~sangjin/2012/12/21/epoll-vs-kqueue.html http://www.eecs.berkeley.edu/~sangjin/2012/12/21/epoll-vs-kq...
- cbsmith 12y agoYeah, it's the opposite, though there are other mechanisms that could do better than epoll, but they need more refinement.
- donavanm 12y agoActually o_async is, imo, broken on linux. Forcing it to direct precludes the use of readahead or the page cache.
- cbsmith 12y agoYeah, O_ASYNC on Linux kinda sucks.