Very Slow Network Transfer

dejan.b · 10-14-2008, 10:51 AM

Hi,

I recently installed Ubuntu Server 8.04 on DELL PowerEdge Server. It works fine, except that the transfer rate is around 2,000 kbytes/s and then sometime it drops to a crawl: 250kbytes/s. So, I tried various things and in the end I changed the network card. The first few seconds was great, more than 9000kbytes/s but then the transfer dropped to around 2500kbytes/s. All the tests were done with only my PC connected to that server. The transfer starts OK with > 9500 but then I can see it kind of "chokes" and it drops down.

I thought that the transfer speed has something to do with UDMA.

hdparm shows:

Code:

# hdparm /dev/sda

/dev/sda:
 IO_support    =  0 (default) 
16-bit)
 HDIO_GET_UNMASKINTR failed: Inappropriate ioctl for device
 HDIO_GET_DMA failed: Inappropriate ioctl for device
 HDIO_GET_KEEPSETTINGS failed: Inappropriate ioctl for device
 readonly      =  0 (off)
 readahead     = 256 (on)
 geometry      = 60801/255/63, sectors = 976773168, start = 0

So, I can't set DMA mode, I browsed the internet on this "Inappropriate ioctl for device" and I haven't found a solution

If that's the problem then the speed of my drive is very low. The speed test shows:

Code:

# hdparm -Tt /dev/sda

/dev/sda:
 Timing cached reads:   786 MB in  2.00 seconds = 392.32 MB/sec
 Timing buffered disk reads:  150 MB in  3.01 seconds =  49.85 MB/sec

That seems OK since the newer server is about %20 faster.

Now, I'm not sure how to handle this. What is causing the packets to be dropped? How can I see what is going on here? mii-diag says everything is fine. Here is the output of lshw -C network:

Code:

# lshw -C network
  *-network               
       description: Ethernet interface
       product: 82541PI Gigabit Ethernet Controller
       vendor: Intel Corporation
       physical id: 4
       bus info: pci@0000:00:04.0
       logical name: eth0
       version: 05
       serial: 00:1b:21:24:e8:35
       size: 1GB/s
       capacity: 1GB/s
       width: 32 bits
       clock: 66MHz
       capabilities: pm pcix bus_master cap_list ethernet physical tp 10bt 10bt-fd 100bt 100bt-fd 1000bt-fd autonegotiation
       configuration: autonegotiation=on broadcast=yes driver=e1000 driverversion=7.3.20-k2-NAPI duplex=full firmware=N/A ip=10.0.1.233 latency=32 link=yes mingnt=255 module=e1000 multicast=yes port=twisted pair speed=1GB/s

I thing I have a network problem not DMA problem but what to do?

Thanks,

Dejan

baldy3105 · 10-15-2008, 07:34 AM

These symptoms are typical of duplex mismatch. Its worth checking that both interfaces are coming up as Full Duplex

dejan.b · 10-15-2008, 08:04 AM

Thanks for your answer.

I set the speed of CISCO switch to 100M, but the problem is still there.

mii-tool on the servers says:

eth0: negotiated 100baseTx-FD, link ok

Dejan

baldy3105 · 10-17-2008, 05:01 AM

You MUST have both ends of a 100BaseT segment set as auto, or all ports hard configured with matching duplex, otherwise you WILL get a duplex mismatch.

Based on IEEE 802.3u, it is not possible to manually configure one link partner for 100 Mbps, full-duplex and still autonegotiate to full-duplex with the other link partner. If you attempt to configure one link partner for 100 Mbps, full-duplex and the other link partner for autonegotiation, it results in a duplex mismatch. This is because one link partner autonegotiates and does not see any autonegotiation parameters from the other link partner and defaults to half-duplex.

Approximately half off the "network performance issues" that get escalated me are due to duplex-mismatch.