Messages in this thread | | | Date | Wed, 23 Jun 1999 00:36:19 +0200 (MEST) | From | Ricardo Galli Granada <> | Subject | Re: New buffer/cache small report and questions |
| |
On Tue, 22 Jun 1999, Andrea Arcangeli wrote: > It still has some bugs but less than in 2.3.7. I think it should be faster > than 2.3.7. I am very interested in numbers :)).
Give some days, so customers won't complain...
I realised something strange testing again the database server (explained below).
Is there any kernel related explanation? Or the test is no reliable?
Just for curiosity, I did the same dbase test again among 2.2.9 and 2.2.10.
2.2.9 is a little faster (or better to say, consumes less CPU), about 2%.
I also noticed the biggest differences are in sys and user times. Sys times in 2.2.10 are about 10% larger than in 2.2.9. I did not find the reason in the patch, may be in VFS or TCP changes? All accessed data was loaded to the page cache (about 150 MB in 450MB free RAM).
The database server I am using does an accept on a TCP socket (only one ip address), which in the 80% was established from a client in the same machine (the other 20% from a PII UP running 2.2.9). The server does a a fork for each new connection.
The "signature" of the process is as follows (snapshot from 2.2.10 after 70.000 operations)
PID TTY MAJFLT MINFLT TRS DRS SIZE SWAP RSS SHRD LIB DT COMMAND 138 ? 37 278889 16 44 208 148 60 44 0 4 /usr/sbin/dbserver_sql
total used free shared buffers cached Mem: 517320 500860 16460 25392 96228 365952 -/+ buffers/cache: 38680 478640 Swap: 129004 3632 125372
** Parent Times User: 00:00:02.07 Sys: 00:00:37.47 Total: 00:00:39.54 ** Total Times User: 00:19:04.77 Sys: 00:15:16.19 Total: 00:34:20.96
-- Ricardo Galli
- 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/
| |