Verify Duplex?

[NcCorduan]

I'm trying to diagnose some network connections, and one of the things I'm looking at is whether we're actually connecting at full duplex. I thought I'd hit the jackpot when my new Fluke was only showing a half-duplex connection (while still showing full or half capable on the switch), but then I discovered that the Fluke is just a half-duplex device so that wasn't doing me any good.

When I go into the switches and do SHOW INTERFACE, the switch-ports, at least, look like they're auto-negotiating to full duplex just fine.

Is there anything I can do on the PC side to see if the computers also indicate a full-duplex connection?

Thanks!

 

[bani2311]

Make sure on the computer you set the etherent card to 100Mbps Fullduplex ,this will confirm the speed

 

[jimbopalmer]

Warning! Will Robinson, Warning!

NEVER set only one end of a wire to full duplex, the negoitation protocol (NWAY) used by ethernet does not work at all if only one end is hard wired to full duplex.

I wrote a FAQ on that in this forum

I tried to remain child-like, all I acheived was childish.

 

[NcCorduan]

Well that explains why when my supervisor recommended setting our cards manually to full duplex the performance on the machines went out the window rather than got better.

So do you know of a way, on the PC side, to verify that your machine has autonegotiated to full duplex?

Thanks!

 

[jimbopalmer]

Sadly once you start hard coding them you must do both sides, auto is a much better place to be.

While I Know how to see Duplex on a Mac or unix I do not know a way in Windows. Fluke makes a net tool I have used that looks in line with the connection to show you what each end can do and is doing

http://www.flukenetworks.com/uk/LAN/Handheld+Testers/NetTool/Specifications.htm

In Win3x and 9x I am not sure one CAN write a generalized duplex program, but in 2000, XP and 2003 it ought to be possible, I just do not know of one.

I tried to remain child-like, all I acheived was childish.

 

[NcCorduan]

Yeah, our Fluke shows the line as capable of being used for Full Duplex, but since it's a half-duplex devcice, it can't help us test auto-negotiation.

I've read your FAQ and done a lot more research now, and I'm sure you're correct about NWAY not being truly compatibile with specified Full Duplex media, but I need some hard data (time trials or performance tests) to show this before my supervisor with accept it. He's asking us to do some time testing, but as you know that's not the most reliable way to check duplex, because it's very mcuh dependent on the type of operation you're using for your time-testing.

I'm really surprised there isn't a sniffer or ping-type utility for Windows that can verify duplex of a connection...

 

[jimbopalmer]

Hope these help with the boss

http://beowulf.es.embnet.org/listarchives/linux-tulip/1999/11/0078.html

http://scyld.com/expert/NWay.html

http://www-2.cs.cmu.edu/afs/cs.cmu..../new/daydreamer/AT2500Adapter/INFO/DUPLEX.TXT

http://docs.sun.com/db/doc/806-1053/6jac7o6ab?a=view

http://docs.sun.com/db/doc/805-2804/6j2mcsm8c?a=view

http://www.intel.com/support/network/adapter/6475.htm

http://www.mail-archive.com/freebsd-net@freebsd.org/msg09308.html

http://mail-index.netbsd.org/tech-net/1999/12/08/0004.html

Note that this is by NASA, so rocket sceintists have trouble with this too

"802.3u is *NOT* clear on when ANER_LPAN will appear on the link partner. However, it seems to infer that if BMCR_AUTOEN is *not* set, then ANER_LPAN won't be set on the link partner, either. This is the key for the link partner to do Parallel Detection [a method of determining 10Mb/s or 100Mb/s signaling rate without NWAY]. Parallel detection is, by definition, to never detect full-duplex mode. Basically, if you're not doing NWAY, you shouldn't be doing full-duplex unless you're going to hard-wire the media on BOTH ENDS OF THE CONNECTION."

I tried to remain child-like, all I acheived was childish.

 

[Nettekkie1010]

This may be a guess but I think the reason no-one can figure out how to verify actual speed/duplex in Windows is this: The auto-negotiate feature is supported and controlled via the NIC itself and the driver associated with it. There is no way Microsoft can check this.. it's vendor specific.

The place I would start is the NIC vendor, not Microsoft. Perhaps they provide some diagnostic software to query the NIC and report actual speed/duplex.

I don't mess with it and the best bet on this is to set all workstations to 100/full in the Local Area Connection AND the switch-port. It's the only way to be sure you have no duplex mis-match.

Haveagoodun!
Nettekkie1010

 

[Comstocb]

Easy way to tell. Veiw the error statistics on the port (on the switch) the PC is attached to. If you see no transmit or receive errors, its full duplex. Zero errors means no collisions and that just isn't the case with half duplex when csma/cd is in effect. The switch will report the statistic down at that level. I poked around windows for a bit and found the commands netstat -e and netstat -s return interesting statistics but not of much value when I tested loading my PC with traffic after setting it to 10 half. It appears that the collisions/errors are discarded as they leave the wire and don't get reported accurately were windows has access to gather information (higher up the OSI ref. model). Your either going to need a utility from the NIC manufacturer or trust what you see in the switch it connects to. I am not a big fan of forcing ports and not using auto negotiate. It used to be a problem (older Macs come to mind) but when I have forced ports in the past, the months roll by and new equipment gets connected that isn't forced and the network team chases its tail wondering what gives until someone figures out the port was forced to some value and needs to be returned to auto-negotiate. Just my two cents

Brian

 

[jimbopalmer]

I always prefer Auto except:

if there is a fiber leg, I usually need to control the entire path, switch to PC (or switch to switch)

Auto just checks the electronics, not the wire. For horrible old wire, force 10/half

Apple and Cisco tend to have read the standard different than anyone else, they sometimes do not play well with others. Sun Solaris before 2.51 is a problem at 100 meg, but that was years ago, upgrade.

So I get to be Auto about 95% of the time.

And yes, documenting why you did NOT choose Auto helps the next poor soul.

I tried to remain child-like, all I acheived was childish.