On Fri, 20 Oct 2000, Mark Hahn wrote:
> > This is something that has been bugging me for a while. I notice
> > on my system that during disk write we do much context switching,
> > but not during disk read. Why is that?
>
> bdflush is broken in current kernels. I posted to linux-mm about this,
> but Rik et al haven't shown any interest. I normally see bursts of
> up to around 40K cs/second when doing writes; I hacked a little
> premption counter into the kernel and verified that they're practially
> all bdflush...
Thanks for the info Mark.
My box still isn't really happy with VM in general. It works fine
until I use swap.. then it starts grinding pretty badly. Something
is still amiss. (I tried and failed to figure out what;) Perhaps
this is related.
time make bzImage
test10-pre4
-j 30 single task
real 12m41.040s 6m25.163s
user 6m19.510s 5m58.300s
sys 1m28.160s 0m21.360s
test1-ac22-class
-j 30 single task
real 7m14.076s 6m24.839s
user 6m18.740s 5m59.660s
sys 0m31.180s 0m21.280s
Classzone (beating a dead horse I suppose) consistantly adds under
one minute to single process build times versus over six minutes for
latest kernel. Swap frenzies like the two below flat don't happen in
a classzone build. Neither do extended periods of mostly idle cpu.
This shows that this load _can_ be handled smoothly and efficiently.
The stock VM chokes on this load.. and IMHO it shouldn't.
procs memory swap io system cpu
r b w swpd free buff cache si so bi bo in cs us sy id
29 5 1 0 1956 4044 11924 0 0 51 176 180 222 86 14 0
31 4 1 0 1452 1740 7432 0 0 156 92 322 897 85 15 0
23 7 1 18184 1404 232 3196 12652 18652 8192 4863 3897 35233 22 77 2
6 25 1 43608 77836 300 23136 49352 36844 18683 9271 6405 17727 46 21 33
1 37 0 40616 84384 348 26080 4916 0 1333 0 1148 1425 27 3 70
0 54 0 39524 79108 364 30852 5528 0 1448 0 672 817 0 1 99
5 35 0 39412 75560 388 33240 2280 0 798 0 351 599 0 4 96
0 41 0 39160 70360 536 34008 336 0 761 0 220 549 4 20 76
3 48 0 38432 70616 540 34608 1264 0 323 0 227 282 2 6 92
1 50 0 37868 66648 760 35536 1260 0 464 0 596 790 4 3 93
2 28 0 37864 65816 832 35776 0 0 133 0 202 264 3 3 94
4 27 0 37860 63880 1008 36304 0 0 221 0 226 340 25 5 70
26 28 0 37732 58888 1120 37148 300 0 403 0 362 538 34 5 61
29 9 0 36844 26040 1176 37788 388 0 513 0 470 940 79 21 0
31 0 0 36828 17332 1200 38184 88 0 112 0 184 281 87 13 0
29 5 0 36100 8740 1212 38056 300 0 173 0 235 482 91 9 0
30 0 0 35420 1940 1220 37348 116 0 276 0 187 328 77 23 0
30 1 0 31424 1460 1224 29584 420 0 129 241 287 531 85 15 0
31 1 0 27088 1460 1232 22292 0 0 115 0 122 378 89 11 0
30 2 0 24916 1464 1232 16296 320 0 112 0 151 412 89 11 0
31 3 0 24664 1952 832 11580 428 1776 190 444 224 460 77 23 0
procs memory swap io system cpu
r b w swpd free buff cache si so bi bo in cs us sy id
30 3 2 50348 1412 228 10424 13532 27664 8013 7097 3660 35359 26 66 8
7 15 1 56140 93120 248 22380 67536 32272 22272 8143 9132 16245 39 7 53
37 15 0 50168 79500 364 29184 6912 0 2649 0 1969 2716 2 3 95
0 13 0 48596 74052 676 33028 5136 0 1446 0 2074 2332 1 3 96
2 28 0 48556 72440 820 33432 92 0 444 0 218 302 6 3 92
7 32 0 45596 46476 892 36980 4780 0 1555 0 1631 2330 9 6 86
28 2 0 45304 40376 992 37200 132 0 139 0 226 876 49 11 41
29 2 0 45304 32432 1072 37840 0 0 180 0 248 1154 81 16 4
25 6 0 45300 23664 1116 38004 4 0 53 0 146 261 90 10 0
31 0 0 45300 15108 1144 38448 0 0 118 0 201 394 93 7 0
31 0 0 45300 7092 1172 38492 0 0 18 82 130 310 96 4 0
29 4 0 43404 1940 1172 36676 0 0 27 188 177 311 95 5 0
30 0 0 37248 1940 1172 29172 0 0 10 0 134 337 93 7 0
30 0 0 31344 1460 1172 22668 0 0 0 0 103 341 93 7 0
30 2 0 26044 1460 1172 16188 0 12 0 3 105 258 97 3 0
30 2 1 24220 1944 1124 9772 0 28 0 7 104 246 82 18 0
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Mon Oct 23 2000 - 21:00:17 EST