Messages in this thread | | | Date | Thu, 20 May 1999 03:04:53 +0200 (CEST) | From | Andrea Arcangeli <> | Subject | 2.3.3_andrea2 & 2.2.9_andrea1 [was Re: Bad apache perfomance wtih linux SMP] |
| |
On Mon, 17 May 1999, Juergen Schmidt wrote:
>I'm measuring plain HTTP-GET on a static html-file with 8 >(Linux-)clients each running up to 64 processes (for a total of 512) >doing HTTP-GET-requests in a tight loop. All files come from a partition >on the RAID-array (RAID 5) which is used for logging, too. File size is >4KByte (I measured 1k and 8k with similar results).
I did a new 2.2.9_andrea1.bz2 just for you. I would be glad if you would try to benchmark it and compare it with the stock 2.2.9 kernel for web-serving or everything else purposes (mainly in SMP). There's all my new VM/buffers/SMP-scheduler/simple-wake-one/etc..etc.. stuff included (NOTE: the unlock_kernel-during-memcpy thing suggested by Andi in an email of this thread is _not_ included in my patch, so you may want to add it by hand in a _secondary_ benchmark stage).
If you'll run a benchmark also over 2.3.x you may give a try also to 2.3.3_andrea2.bz2 (that is the equivalent of 2.2.9_andrea1 but against the unstable branch).
To compile it just apply it and run `make oldconfig'. Say N for the kernel lock profiling (to not lose one bit of performances in getting a bit more of info). Then make dep clean and go ahead.
Thanks :-).
ftp://e-mind.com/pub/andrea/kernel/2.2.9_andrea1.bz2 ftp://e-mind.com/pub/andrea/kernel/2.3.3_andrea2.bz2
As usual there are the two mirrors of the /pub/andrea sub-tree at:
ftp://ftp.linux.it/pub/People/andrea/ (Italy, Thanks to linux.it guys)
ftp://master.softaplic.com.br/pub/andrea/ (Brazil, Thanks to Edesio Costa e Silva <edesio@acm.org>, 2MBits/sec)
Andrea Arcangeli
PS. Don't forget to make sure that apache is sleeping in accept(2) and _not_ in flock(2), just run strace -p `pid of one apache child` to know where it's sleeping. If it's sleeping in flock(2) you must recompile it with the option to sleep in accept.
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |