Rik spreading bullshit about VM
About running the Andrea VM in 4M:
I booted in single user mode.
bash-2.05a# uname -a
Linux mountain 2.4.18pre2aa2 #1 Wed Jan 9 21:44:03 EST 2002 i586 unknown
bash-2.05a# dmesg| grep mem
Kernel command line: BOOT_IMAGE=2418p2aa2 ro root=1602 console=ttyS1,38400n8 single mem=4m
Memory: 2100k/4096k available (891k kernel code, 1608k reserved, 215k data, 196k init, 0k highmem) Freeing unused kernel memory: 196k freed
Manually started syslogd, klogd and brought up the lo and eth0 interfaces and started sshd. I can ssh into the box:
mountain:/$ ps aux USER PID %CPU %MEM VSZ RSS TTY STAT START TIME COMMAND root 1 0.4 0.0 1288 0 ? SW 19:24 0:06 init [S] root 2 0.0 0.0 0 0 ? SW 19:24 0:00 [keventd] root 3 0.0 0.0 0 0 ? SWN 19:24 0:00 [ksoftirqd_CPU0] root 4 0.1 0.0 0 0 ? SW 19:24 0:02 [kswapd] root 5 0.0 0.0 0 0 ? SW 19:24 0:00 [bdflush] root 6 0.0 0.0 0 0 ? SW 19:24 0:00 [kupdated] root 7 0.0 0.0 0 0 ? SW 19:24 0:00 [kreiserfsd] root 21 0.0 0.0 1288 0 ttyS1 SW 19:25 0:00 init [S] root 22 0.0 0.0 2212 0 ttyS1 SW 19:25 0:00 bash root 46 0.0 0.0 1444 0 ? SW 19:37 0:00 /usr/sbin/syslogd -m0 root 49 0.0 0.0 1292 0 ? SW 19:38 0:00 /usr/sbin/klogd -c3 -x -k /boot/System.map-2.4.18pre2aa2 root 63 0.0 0.0 0 0 ? SW 19:47 0:00 [eth0] root 68 0.5 0.0 2672 0 ? SW 19:47 0:00 sshd root 69 0.9 0.8 2756 20 ? D 19:48 0:00 sshd hrandoz 70 1.9 0.0 2192 0 pts/0 SW 19:48 0:01 -bash hrandoz 75 40.0 4.1 2504 96 pts/0 R 19:49 0:00 ps aux
mountain:/$ cat /proc/meminfo
total: used: free: shared: buffers: cached:
Mem: 2351104 2232320 118784 0 344064 249856
Swap: 156270592 1409024 154861568
MemTotal: 2296 kB
MemFree: 116 kB
MemShared: 0 kB
Buffers: 336 kB
Cached: 204 kB
SwapCached: 40 kB
Active: 68 kB
Inactive: 516 kB
HighTotal: 0 kB
HighFree: 0 kB
LowTotal: 2296 kB
LowFree: 116 kB
SwapTotal: 152608 kB
SwapFree: 151232 kBhrandoz@mountain:/$ /sbin/swapon -s Filename Type Size Used Priority /dev/hda3 partition 152608 1372 -1
top:
7:55pm up 31 min, 1 user, load average: 1.30, 0.71, 0.32
16 processes: 15 sleeping, 1 running, 0 zombie, 0 stopped
CPU states: 0.1% user, 4.6% system, 0.0% nice, 95.1% idle Mem: 2296K av, 2096K used, 200K free, 0K shrd, 316K buff Swap: 152608K av, 1488K used, 151120K free 96K cached
PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME COMMAND
79 hrandoz 14 0 316 200 184 R 2.0 8.7 0:02 top
69 root 18 0 352 0 0 SW 1.9 0.0 0:03 sshd
1 root 20 0 56 0 0 SW 0.0 0.0 0:06 init
2 root 20 0 0 0 0 SW 0.0 0.0 0:00 keventd
3 root 20 19 0 0 0 SWN 0.0 0.0 0:00 ksoftirqd_CPU0
4 root 20 0 0 0 0 SW 0.0 0.0 0:03 kswapd
5 root 3 0 0 0 0 SW 0.0 0.0 0:00 bdflush
6 root 20 0 0 0 0 SW 0.0 0.0 0:00 kupdated
7 root 20 0 0 0 0 SW 0.0 0.0 0:00 kreiserfsd
21 root 9 0 56 0 0 SW 0.0 0.0 0:00 init
22 root 20 0 296 0 0 SW 0.0 0.0 0:00 bash
46 root 20 0 108 0 0 SW 0.0 0.0 0:00 syslogd
49 root 20 0 72 0 0 SW 0.0 0.0 0:00 klogd
63 root 20 0 0 0 0 SW 0.0 0.0 0:00 eth0
68 root 20 0 276 0 0 SW 0.0 0.0 0:00 sshd
70 hrandoz 20 0 276 0 0 SW 0.0 0.0 0:01 bashbash-2.05a# df -kT
Filesystem Type 1k-blocks Used Available Use% Mounted on
/dev/hdc2 reiserfs 9768728 4194328 5574400 43% / /dev/hdc3 ext2 8064432 4912 7649872 1% /opt/testing
Running 4M is interesting. If I really wanted to run 4M, I'd use ash for the shell, and try uClibc instead of glibc.
The test that really impresses me the most about 2.4.18pre2aa2 is what I see on this test on an Athlon 1333 with 1024MB RAM:
simultaneously:
run continuous loop of mtest01 -p 85 -w # allocate and write to 85% of VM create and cpio 10 330 MB files
nice -19 setiathome &
listen to mp3blaster # a few skips at the beginning, then smooth.
I've tested a bunch of kernels lately. This is only what's sitting in /boot since I last cleaned it out:
mountain:/boot$ ls vml* vmlinuz vmlinuz-2.4.17rc2aa2-old vmlinuz-2.4.18-pre3 vmlinuz-2.4.18pre2 vmlinuz-2.4.18pre3ll vmlinuz-2.5.1-dj11 vmlinuz-2.5.2-pre10 vmlinuz-2.5.2-pre9 vmlinuz-2.4.17 vmlinuz-2.4.17rc2aa2-wli vmlinuz-2.4.18-pre4 vmlinuz-2.4.18pre2aa1 vmlinuz-2.4.18pre3pe vmlinuz-2.5.1-dj13 vmlinuz-2.5.2-pre11 vmlinuz-2.5.2-pre9mingo vmlinuz-2.4.17-rmap11a vmlinuz-2.4.17rmap11b vmlinuz-2.4.18pre1-mjc2nio vmlinuz-2.4.18pre2aa2 vmlinuz-2.4.18pre3pelb vmlinuz-2.5.1-dj14 vmlinuz-2.5.2-pre5 vmlinuz.old vmlinuz-2.4.17rc2aa2 vmlinuz-2.4.18-pre1 vmlinuz-2.4.18pre1mjc2 vmlinuz-2.4.18pre3-ac2 vmlinuz-2.5.1 vmlinuz-2.5.2 vmlinuz-2.5.2-pre6-mingo
IMHO, 2.4.18pre2aa2 is the best!
Some kernel test results at:
http://home.earthlink.net/~rwhron/kernel/k6-2-475.html
# from the 4M box mountain:/boot$ uptime 8:24pm up 1:00, 1 user, load average: 0.08, 0.04, 0.09
Signature · 1 line
Randy Hron
In article <[email protected]> you write:
| About running the Andrea VM in 4M: | I booted in single user mode. | | bash-2.05a# uname -a | Linux mountain 2.4.18pre2aa2 #1 Wed Jan 9 21:44:03 EST 2002 i586 unknown | | bash-2.05a# dmesg| grep mem | Kernel command line: BOOT_IMAGE=2418p2aa2 ro root=1602 console=ttyS1,38400n8 single mem=4m | Memory: 2100k/4096k available (891k kernel code, 1608k reserved, 215k data, 196k init, 0k highmem) | Freeing unused kernel memory: 196k freed
[...snip...]
| I've tested a bunch of kernels lately. This is only what's sitting in /boot | since I last cleaned it out: | | mountain:/boot$ ls vml* | vmlinuz vmlinuz-2.4.17rc2aa2-old vmlinuz-2.4.18-pre3 | vmlinuz-2.4.18pre2 vmlinuz-2.4.18pre3ll vmlinuz-2.5.1-dj11 | vmlinuz-2.5.2-pre10 vmlinuz-2.5.2-pre9 vmlinuz-2.4.17 | vmlinuz-2.4.17rc2aa2-wli vmlinuz-2.4.18-pre4 vmlinuz-2.4.18pre2aa1 | vmlinuz-2.4.18pre3pe vmlinuz-2.5.1-dj13 vmlinuz-2.5.2-pre11 | vmlinuz-2.5.2-pre9mingo vmlinuz-2.4.17-rmap11a vmlinuz-2.4.17rmap11b | vmlinuz-2.4.18pre1-mjc2nio vmlinuz-2.4.18pre2aa2 vmlinuz-2.4.18pre3pelb | vmlinuz-2.5.1-dj14 vmlinuz-2.5.2-pre5 vmlinuz.old | vmlinuz-2.4.17rc2aa2 vmlinuz-2.4.18-pre1 vmlinuz-2.4.18pre1mjc2 | vmlinuz-2.4.18pre3-ac2 vmlinuz-2.5.1 vmlinuz-2.5.2 | vmlinuz-2.5.2-pre6-mingo | | IMHO, 2.4.18pre2aa2 is the best!
Of the current kernels it may well be, I have tried it on two largish
machines and it worked "right," I'm building it on a small and slow machine to try that. However, I was not happy with the default bdflush settings, and people who don't see what they want should use the expanded capabilities of -aa before complaining. Also, the performance I've seen so far, and I have NOT run a full set of tests, would indicate that it is slightly better than 2.4.13-acN (N is 5, 7 or 8, don't have it on this machine).
I am looking forward to testing on many configs, and just for the
extra tuning tools in bdflush I think it will be good on all of them. As I posted before, I think the best scheduler is the one which doesn't have "jackpot cases" which produce really bad performance. With a little tuning I believe -aa is there.
Finally, that said I'm trying a patch of my own to -aa, which is why
I haven't run the full set of tests, there are two things I think will make it even better, and I am finally getting to understand the code, little as I wanted to.
I can't disagree with Rik on the VM in the unpatched kernel for
several months. It really was not good, and both the -ac and -aa kernels were taking aim at that problem. IMHO the changes went in before the bugs went out. Needless to say Rik should not apply for a job as a diplomat, but I can't disagree with the existance of a problem. If RH uses a custom VM he was factual about that, although there may be several reasons for the choice.
Maybe we could deflect the pissing contest back to technical
discussion now?
Signature · 3 lines
bill davidsen <[email protected]> CTO, TMR Associates, Inc Doing interesting things with little computers since 1979.
Quoted 5 lines · bill davidsen
I can't disagree with Rik on the VM in the unpatched kernel for
several months. It really was not good, and both the -ac and -aa kernels were taking aim at that problem. IMHO the changes went in before the bugs went out. Needless to say Rik should not apply for a job as a diplomat, but I can't disagree with the existance of a problem.
Good thing I'm a programmer, I'm allowed to say ambiguous stuff on IRC. If I were a diplomat, I'd only be allowed to unambiguously say nothing.
Quoted 2 lines
If RH uses a custom VM he was factual about that, although there may be several reasons for the choice.
I've tried to be as factual as possible; I've also learnt that articles and IRC really use a different kind of language ... some of the things I said turned out pretty ambiguous ;)
Quoted 2 lines
Maybe we could deflect the pissing contest back to technical
discussion now?
I've released rmap-11c today ;)
cheers,
Rik
Signature · 4 lines
"Linux holds advantages over the single-vendor commercial OS"
-- Microsoft's "Competing with Linux" document
http://www.surriel.com/ http://distro.conectiva.com/Excellent, Rik!
I've posted a diff for rmap-11c against vanilla 2.4.18-pre4 at
http://www.cs.unc.edu/~chenda/Other/2.4.18-pre4_to_rmap11c.diff
Quoted 4 lines · Rik van Riel
Quoted 2 lines
Maybe we could deflect the pissing contest back to technical
discussion now?I've released rmap-11c today ;)
Signature · 2 lines
Dan Chen [email protected] GPG key: www.unc.edu/~crimsun/pubkey.gpg.asc