P.S. It seems SLAB/SLUB doesn't make much difference, just seems SLUB use a bit more some resources. Still problem appearing with softirqd/0 or softirqd/1 taking a lot CPU (while on 2.6.21 it never came up of the "top"), and doesn't matter - if NAPI on e1000 disabled or enabled. Additionally disabling all disk operations doesn't make any difference. defaulthost ~ #mpstat 1 Linux 2.6.23-rc8-build-0017 (defaulthost) 09/30/07 21:20:29 CPU %user %nice %sys %iowait %irq %soft %steal %idle intr/s 21:20:30 all 21.11 0.00 14.57 8.04 0.00 48.74 0.00 7.54 3632.00 21:20:31 all 23.38 0.00 12.94 10.95 0.00 52.74 0.00 0.00 1524.00 21:20:32 all 21.50 0.00 16.50 8.50 0.00 53.50 0.00 0.00 1473.00 21:20:33 all 22.39 0.00 14.93 4.98 0.50 53.23 0.00 3.98 1492.08 21:20:34 all 5.31 0.00 0.97 0.48 0.00 93.24 0.00 0.00 1351.96 On 2.6.21 it looks: defaulthost ~ #mpstat 1 Linux 2.6.21-gentoo-r3-build-0015 (defaulthost) 09/30/07 18:21:50 CPU %user %nice %sys %iowait %irq %soft %steal %idle intr/s 18:21:51 all 29.50 0.00 21.50 12.50 0.00 6.00 0.00 30.50 8416.00 18:21:52 all 26.00 0.00 19.00 12.50 0.50 7.00 0.00 35.00 8663.64 18:21:53 all 27.23 0.00 17.82 13.37 0.99 6.44 0.00 34.16 7828.71 18:21:54 all 26.13 0.00 16.08 18.59 0.00 7.04 0.00 32.16 8657.58 18:21:55 all 22.89 0.00 16.92 18.91 0.50 6.47 0.00 34.33 7957.43 18:21:56 all 28.14 0.00 21.61 17.59 0.50 6.03 0.00 26.13 8531.00 -- Denys Fedoryshchenko Technical Manager Virtual ISP S.A.L. -
| Mark Lord | 2.6.25-rc8: FTP transfer errors |
| Kamalesh Babulal | Re: 2.6.23-rc6-mm1 |
| Greg Kroah-Hartman | [PATCH 025/196] paride: Convert from class_device to device for block/paride |
| Stephen Rothwell | Announce: Linux-next (Or Andrew's dream :-)) |
git: | |
| Linus Torvalds | Re: iptables very slow after commit 784544739a25c30637397ace5489eeb6e15d7d49 |
| David Miller | Re: [GIT]: Networking |
| Gerrit Renker | [PATCH 18/37] dccp: Support for Mandatory options |
| Jarek Poplawski | [PATCH] pkt_sched: Destroy gen estimators under rtnl_lock(). |
