[Q] poor network performance

[sirola at fisica.unige.it]

Hello,

I have Dell PE1750 server with FreeBSD 4.10-RELEASE-p2 freshly
installed. The server shows very low transfer rates with other hosts, 
but i can't see any error/strange data on interfaces/mbufs
statistics. I tried scp on both interfaces and it's very slow (approx
100KBytes/s), while if it try to route back the response packets (acks) on
the other interface it goes very fast (100MBytes/s). Swapping cables
and configuration doesn't show any difference.

The host is configured as follows:

* two gigabit ethernets: <Broadcom BCM5704C Dual Gigabit Ethernet, ASIC rev. 0x2002> 

seen as bge0 and bge1
bge0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> mtu 1500
        options=3<RXCSUM,TXCSUM>
        inet 192.168.2.3 netmask 0xffffff00 broadcast 192.168.2.255
        inet6 fe80::20b:dbff:fe93:3b5a%bge0 prefixlen 64 scopeid 0x1
        ether 00:0b:db:93:3b:5a
        media: Ethernet 100baseTX <full-duplex>
        status: active
bge1: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> mtu 1500
        options=3<RXCSUM,TXCSUM>
        inet 192.168.3.3 netmask 0xffffff00 broadcast 192.168.3.255
        inet6 fe80::20b:dbff:fe93:3b5b%bge1 prefixlen 64 scopeid 0x2
        ether 00:0b:db:93:3b:5b
        media: Ethernet 100baseTX <full-duplex>
        status: active

bge1 has an alias to the host public ip address (not shown).

* 1 GB RAM

* sysctl.conf as follows:
  kern.ipc.somaxconn=1024
  kern.ipc.shmmax=700000000
  kern.ipc.shmall=524280
  kern.ipc.shm_use_phys=1

* loader.conf as follows:
  kern.ipc.nmbclusters="32768"

* routing is as follows:
netstat -rn
Routing tables

Internet:
Destination        Gateway            Flags    Refs      Use  Netif Expire
default            192.168.3.1        UGSc        1       64   bge1
127.0.0.1          127.0.0.1          UH          1     2683    lo0
192.168.0          192.168.2.1        UGSc        6     2385   bge0
192.168.2          link#1             UC          2        0   bge0
192.168.2.1        00:01:02:24:da:2c  UHLW        6     8029   bge0    972
192.168.2.3        00:0b:db:93:3b:5a  UHLW        0       87    lo0
192.168.3          link#2             UC          2        0   bge1
192.168.3.1        00:05:1c:0b:26:b3  UHLW        2     6861   bge1    988
192.168.3.3        00:0b:db:93:3b:5b  UHLW        0       17    lo0
<public network>   link#2             UC          1        0   bge1
<public ip>        00:0b:db:93:3b:5b  UHLW        0     6167    lo0


While transfering files, I see what (to me) seems a "normal" behavior
with netstat -m

515/704/131072 mbufs in use (current/peak/max):
        515 mbufs allocated to data
512/596/32768 mbuf clusters in use (current/peak/max)
1368 Kbytes allocated to network (1% of mb_map in use)
0 requests for memory denied
0 requests for memory delayed
0 calls to protocol drain routines

This stuff is going to drive me crazy... 
Thanks in advance for your time / help,
e.

-- 
Enrico Sirola <sirola at fisica.unige.it>