Autonegotiation advice

Readers respond to a Kevin Tolly column on autonegotiation.

As the principal developer of the Clause 28 autonegotiation test system/test suites within the Fast Ethernet and Gigabit Ethernet Consortiums of the University of New Hampshire’s InterOperability Lab, and having been associated with the testing of every piece of autonegotiation-capable silicon since 1996, I must voice a few comments regarding Brian Tolly’s column “Do you believe in the autonegotiation fairy?“

Tolly writes: “After countless hours trying to identify the reason for the extremely low throughput, we discovered an overabundance of frame checksum errors on the counters of the questionable switch. After a clearing of the counters and subsequent runs, it was clear that we had identified our issue. What could cause these errors? A colleague of mine recommended hard coding the speed and duplex. Much to my dismay, it solved our problems.”

For more than seven years, full-duplex Ethernet has existed and the issue of duplex mismatch has been omnipresent. The result of configuring one link partner to full duplex and the other to half duplex always has been catastrophic. There also is the real possibility that the vendor’s drivers/firmware/software simply misconfigure the Ethernet media access control so that, though the MAC should be in half-duplex mode, it is configured to full-duplex mode (or vice versa). This is a driver/firmware/software issue and has little to do with autonegotiation.

Muddying the concept of autonegotiation with articles that are pointing out software issues, as opposed to flaws in autonegotiation, will do nothing to improve network administrators’ confidence in autonegotiation – confidence that must grow as 1000Base-T rolls out and 10GBase-T emerges. Please do not simply shut off autonegotiation. Rather, contact the manufacturer, get a fixed driver, and make the world a better place for us all.

Bob Noseworthy

10-Gigabit Ethernet Consortium Manager

University of New Hampshire InterOperability Lab

Durham, N.H.

None of us believes in the autonegotiation fairy. I have all-too-frequent problems with new 10M/100M bit/sec network interface cards going into 10M/100M bit/sec switched hubs, even to the extent of no connection at all. I suspect that a few scraps of Category 3 cable still hanging around have something to do with this, although when it won’t connect, I can force a 100M bit/sec connect by setting the media.

I work for a securities company. We too have found that autonegotiation causes performance problems and have hard coded the speed and duplex settings into our devices. Tolly does not mention why the flapping between port speeds was occurring. Is it that some vendors did not code the feature properly as the spec details? Are there any timers involved that do not match during the negotiation window?

Herb Strohsahl

Brooklyn, N.Y.

Tolly responds:I am not sure where the problem lies. I have received dozens of e-mails like yours, but none from the vendor community. Vendors: Any answers?

Regarding Mr. Rauch’s comments about Category 3 cable, I would not assume that this is the problem. We have seen some interesting results in our “Gigabit to the desktop” study using mangled, abused, subpar Category 5 cable. Makes us wonder if could run Fast Ethernet speeds over barbed wire!