netdev
[Top] [All Lists]

Re: [Fwd: Re: possible bug x86 2.4.2 SMP in IP receive stack]

To: feldy@xxxxxxxx, kuznet@xxxxxxxxxxxxx
Subject: Re: [Fwd: Re: possible bug x86 2.4.2 SMP in IP receive stack]
From: Bob Felderman <feldy@xxxxxxxx>
Date: Wed, 7 Mar 2001 11:02:41 -0800 (PST)
Cc: andrewm@xxxxxxxxxx, davem@xxxxxxxxxx, netdev@xxxxxxxxxxx, pp@xxxxxxxxxxxxxx
Sender: owner-netdev@xxxxxxxxxxx
=> From kuznet@xxxxxxxxxxxxx Wed Mar  7 10:55:04 2001
=> Actually, the fact that eviction happens during your tests implies
=> that you should increase eviction threshold (ipfrag_high_thresh)
=> to value, when eviction never happens on LAN tests. It should happen
=> only if driver has lost some fragment... not good on LAN too.

Is there a counter telling me when "eviction happens"?


I rebooted both machines to get a clean start and I see 
good performance. The UDP sweet-spot is about 8KB (less than 1 mtu).
666MHz dual-processor P-II machines.


rcc 31% udp_range rcc2-t
UDP UNIDIRECTIONAL SEND TEST to rcc2-t
Socket  Message  Elapsed      Messages                
Size    Size     Time         Okay Errors   Throughput
bytes   bytes    secs            #      #   10^6bits/sec

2097152   32768   10.00       62720      0    1644.89
2097152           10.00        7780            204.04

2097152   16384   11.00      119694      0    1426.33
2097152           11.00       19009            226.52

2097152    8192   6.86       201220      0    1922.43
2097152           6.86       166168           1587.55

2097152    4096   5.48       335840      0    2008.92
2097152           5.48       231505           1384.81

2097152    2048   9.26       504164      0     891.66
2097152           9.26       329154            582.14

2097152    1024   9.52       697557      0     600.01
2097152           9.52       517130            444.81

2097152     512   13.77     1146536      0     341.11
2097152           13.77      639733            190.33

2097152     256   11.00     1451247      0     270.30
2097152           11.00      640510            119.30

2097152     128   9.00      1432713      0     163.08
2097152           9.00       605639             68.94

2097152      64   4.73      1641784      0     177.88
2097152           4.73       625544             67.77

2097152      32   3.58      1436986      0     102.77
2097152           3.58       635321             45.44

2097152      16   9.99      1650593      0      21.14
2097152           9.99       642957              8.23

2097152       8   9.99      1410515      0       9.03
2097152           9.99       644561              4.13

2097152       4   6.36      1657061      0       8.33
2097152           6.36       643897              3.24




<Prev in Thread] Current Thread [Next in Thread>