Slowish 5.3 network throughput (LAN)

[Nelis Lamprecht]

On Thu, 11 Nov 2004 14:38:48 -0600 (CST), klr at 6s-gaming.com
<klr at 6s-gaming.com> wrote:
> Hi list,
> 
> About 4 days ago I downloaded 5.3-RELEASE (.iso) from ftp.nl. at about
> 6mB/s from a server I have on a .nl provider. Today I wanted to do some
> testing, and while downloading that same ISO (which I downloaded at 6
> megabytes/s) from two different servers on the same subnet, the best I got
> was 130kB/s.
> 
> My outgoing port 80 pipe on the server allows up to 40mbits, so this rate
> is incredibly low, and the servers are all on the same subnet!
> 
> The clients are downloading the .iso from an apache 1.3.33 server with a
> basic configuration. I didn't have time to test over NFS, to check if it
> would be any faster..
> 
> What's happening here? I can download at 1mBps from my crappy P200 MMX w/
> freebsd 4.10 at my lan. This server is a Dual Xeon 2.4Ghz w/ 2GB ram and a
> decent hdd. It should saturate the 100mbps, but won't go past 130kB/s.
> 
> Could this be because the server NIC is an em(4) ? I heard there are some
> problems with the em driver under 5.3.
> 
> Here are some specs:
> 
> CPU: Intel(R) Xeon(TM) CPU 2.40GHz (2399.33-MHz 686-class CPU)
>   Hyperthreading: 2 logical CPUs
> FreeBSD/SMP: Multiprocessor System Detected: 4 CPUs
> SMP: AP CPU #3 Launched!
> SMP: AP CPU #1 Launched!
> SMP: AP CPU #2 Launched!
> 
> em0: <Intel(R) PRO/1000 Network Connection, Version - 1.7.35> port
> 0x3000-0x303f mem 0xfc200000-0xfc21ffff irq 54 at device 3.0 on pci2
> 
> 487 mbufs in use
> 270/32768 mbuf clusters in use (current/max)
> 0/4/6656 sfbufs in use (current/peak/max)
> 661 KBytes allocated to network
> 0 requests for sfbufs denied
> 0 requests for sfbufs delayed
> 0 requests for I/O initiated by sendfile
> 28 calls to protocol drain routines
> 
> FreeBSD celestia.celeritystorm.com 5.3-RELEASE FreeBSD 5.3-RELEASE #0: Mon
> Nov 1 22:21:19 UTC 2004
> klr at celestia.celeretystorm.com:/usr/obj/usr/src/sys/celestia i386
> 
> I don't have polling enabled. sacks on, delayed acks on.
> 
> the PF queue:
> 
> queue  http_out bandwidth 40Mb priority 5
>   [ pkts:      59257  bytes:   88471020  dropped pkts:      0 bytes:      0 ]
>   [ qlength:   0/ 50  borrows:      0  suspends:    177 ]
> 
> Any ideas of what might be causing the tremendous slowdown ?
> 

We have several machines which use the em driver but I haven't used
5.3 with them yet. What I did notice though that when I set the
managed switch and the interface to full duplex instead of letting it
auto sense it made a noticable difference in speed. I would be
interested to know if 5.3 has the same effect.

ifconfig em0 media 100baseTX mediaopt full-duplex

Nelis