Re: 2.4.19rc2aa1

From: J.A. Magallon (jamagallon@able.es)
Date: Fri Jul 19 2002 - 19:10:41 EST


On 2002.07.20 "Feldman, Scott" wrote:
>Jamagallon wrote:
>
>> >diff between 2.4.19rc1aa2 and 2.4.19rc1aa2:
>> >
>> >Only in 2.4.19rc1aa2: 000_e100-2.0.30-k1.gz
>> >Only in 2.4.19rc2aa1: 000_e100-2.1.6.gz
>> >Only in 2.4.19rc1aa2: 000_e1000-4.2.17-k1.gz
>> >Only in 2.4.19rc2aa1: 000_e1000-4.3.2.gz
>> >
>
>>More on this.
>
>>We have two interfaces:
>>04:04.0 Ethernet controller: Intel Corp. 82557 [Ethernet Pro 100] (rev 08)
>03:01.0 Ethernet
>>controller: Intel Corp. 82543GC Gigabit Ethernet Controller (rev 02)
>
>>NetPipe (tcp) shows numbers like 80Mb/s for e100 and 500Mb/s for e1000. So
>efficiency is much >much higher for e100 driver+card than e1000. I have to
>dig, perhaps e100 is doing zerocopy and >e1000 is not ?
>
>>Any ideas ?
>
>If e100 is sending from the zerocopy path, e1000 is doing the same.
>

e100.txt:
   - Support for Zero copy on 82550-based adapters. This feature provides
     faster data throughput and significant CPU usage improvement in systems
     that use the relevant system call (sendfile(2)).
  (does this include the on-board 82557, not even listed in e100.txt ? )

e1000.txt
   - Zero copy. This feature provides faster data throughput. Enabled by
     default in supporting kernels. It is not supported on the Intel(R)
     PRO/1000 Gigabit Server Adapter. (==82542)
  (so I

>There are several factors that may be limiting your throughput on e1000.
>Assuming you have enough CPU umph and bus bandwidth, and your netpipe link
>partner and switch are willing, you should be able to approach wire speed.
>

Master/sender:
        Dual P4Xeon 1.8GHz, Pro/1000 T Server Board (64bit slot, 66MHz)
Slave/receiver:
        Dual PIII 1GHz, same board, same slot
Switch:
        Intel(R) NetStructure(TM) 470T

netpipe over e100:
Node receiver...
Master transmitter...
Latency: 0.000057
Now starting main loop
  0: 4096 bytes 7 times --> 24.96 Mbps in 0.001252 sec
  1: 8192 bytes 7 times --> 46.45 Mbps in 0.001345 sec
  2: 12288 bytes 92 times --> 46.03 Mbps in 0.002037 sec
  3: 16384 bytes 81 times --> 58.32 Mbps in 0.002143 sec
  4: 20480 bytes 87 times --> 56.26 Mbps in 0.002777 sec
  5: 24576 bytes 72 times --> 71.59 Mbps in 0.002619 sec
  6: 28672 bytes 79 times --> 62.76 Mbps in 0.003486 sec
  7: 32768 bytes 61 times --> 75.23 Mbps in 0.003323 sec
  8: 36864 bytes 65 times --> 66.52 Mbps in 0.004228 sec
  9: 40960 bytes 52 times --> 77.20 Mbps in 0.004048 sec
 10: 45056 bytes 55 times --> 69.32 Mbps in 0.004959 sec
 11: 49152 bytes 45 times --> 74.98 Mbps in 0.005002 sec
 12: 53248 bytes 45 times --> 71.41 Mbps in 0.005689 sec
 13: 57344 bytes 40 times --> 76.24 Mbps in 0.005738 sec
 14: 61440 bytes 40 times --> 73.03 Mbps in 0.006419 sec
 15: 65536 bytes 36 times --> 77.13 Mbps in 0.006483 sec
 16: 69632 bytes 36 times --> 74.10 Mbps in 0.007170 sec
 17: 73728 bytes 32 times --> 78.04 Mbps in 0.007208 sec
 18: 77824 bytes 32 times --> 75.42 Mbps in 0.007872 sec
 19: 81920 bytes 30 times --> 78.61 Mbps in 0.007950 sec
 20: 86016 bytes 29 times --> 76.45 Mbps in 0.008584 sec

(around 75Mb/s, 75% of bandwidth)

netpipe over e1000:
Node receiver...
Master transmitter...
Latency: 0.000058
Now starting main loop
  0: 4096 bytes 7 times --> 204.44 Mbps in 0.000153 sec
  1: 8192 bytes 7 times --> 303.61 Mbps in 0.000206 sec
  2: 12288 bytes 607 times --> 361.83 Mbps in 0.000259 sec
  3: 16384 bytes 643 times --> 408.76 Mbps in 0.000306 sec
  4: 20480 bytes 613 times --> 424.47 Mbps in 0.000368 sec
  5: 24576 bytes 543 times --> 458.94 Mbps in 0.000409 sec
  6: 28672 bytes 509 times --> 474.85 Mbps in 0.000461 sec
  7: 32768 bytes 465 times --> 491.99 Mbps in 0.000508 sec
  8: 36864 bytes 430 times --> 443.68 Mbps in 0.000634 sec
  9: 40960 bytes 350 times --> 448.35 Mbps in 0.000697 sec
 10: 45056 bytes 322 times --> 455.34 Mbps in 0.000755 sec
 11: 49152 bytes 301 times --> 464.07 Mbps in 0.000808 sec
 12: 53248 bytes 283 times --> 464.18 Mbps in 0.000875 sec
 13: 57344 bytes 263 times --> 467.06 Mbps in 0.000937 sec
 14: 61440 bytes 247 times --> 476.95 Mbps in 0.000983 sec
 15: 65536 bytes 237 times --> 482.73 Mbps in 0.001036 sec
 16: 69632 bytes 226 times --> 488.26 Mbps in 0.001088 sec
 17: 73728 bytes 216 times --> 473.16 Mbps in 0.001189 sec
 18: 77824 bytes 198 times --> 472.22 Mbps in 0.001257 sec
 19: 81920 bytes 188 times --> 481.13 Mbps in 0.001299 sec
 20: 86016 bytes 182 times --> 478.84 Mbps in 0.001371 sec

(peak at 491, not even 50% bandwith...)

I have not played with sysctl:
annwn:/proc/sys/net/ipv4> cat tcp_wmem
4096 16384 131072
annwn:/proc/sys/net/ipv4> cat tcp_rmem
4096 87380 174760

Something can be limiting bandwith in the switch ???

TIA

-- 
J.A. Magallon             \   Software is like sex: It's better when it's free
mailto:jamagallon@able.es  \                    -- Linus Torvalds, FSF T-shirt
Linux werewolf 2.4.19-rc2-jam1, Mandrake Linux 8.3 (Cooker) for i586
gcc (GCC) 3.1.1 (Mandrake Linux 8.3 3.1.1-0.8mdk)
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/



This archive was generated by hypermail 2b29 : Tue Jul 23 2002 - 22:00:32 EST