Gigabit throughput tests?

Is there some reason you're asking a question about wiring in a wireless newsgroup? Not a problem, but I'm curious as to your logic.

I klike IPerf. Versions available for most OS's:

Setup a known good machine running Iperf as a server. Run: iperf -s On the various clients, run: iperf -c ip_address_of_server There are lots of other options to play with. Run the test in both directions with: iperf -r -c ip_address_of_server as creative wiring sometimes causes traffic to work well in one direction, but not the other.

As for watching large files go past, I kinda like to watch the flashing lights on the switch. They're kinda hypnotic and counteract the effects of coffee and running wires all night. If you want detail, setup a sniffer computer, arrange a monitor port or use an ethernet tap (or hub), and use Ethereal or WireShark, with a filter, to monitor the traffic.

Doh! Just due to cross-posting blindness. I have these 2 ng's as posting defaults and I have to *remember* to delete the inappropriate ones. As I have done now...

Thanks, Jeff.

might try netperf -

John E. wrote in part:

Personally I like ole'grandad, `ttcp` .

But be aware: very few machines have the capability of saturating GBE. The PCI bus doesn't have the bandwidth. The best many can do is aroun 300 Mbps.

Consider asking your cabling contractor for Cat5e certification tests. These are the standard when cabling reliability is important.

-- Robert

Thanks. I'm really glad to know that. Keeps expectations realistic.

I'm that guy.

What equipment is used to certify C5e?

John E. wrote in part:

Fluke Pentascanner or equivalent. A 5 k$ device. Should measure length by TDR, NEXT, FEXT, ELFNEXT

-- Rpbert

Very few _old_ machines perhaps, but I suspect that apart perhaps from systems built really on the cheap, that anything built withing the last three years has HW which aught to be capable of gigabit ethernet link-rate. IIRC it was only "PCI 1x" aka 32-bit 33 MHz which didn't have the oomph for gigabit ethernet. Certainly PCI 4x (64 bit 66 MHz) had enough oomph, and any PCI-X or PCIe aught to these days.

rick jones

Rick Jones wrote in part:

You would think so, but I wouldn't automatically assume a consumer-level machine has been designed with the gigabit chip hanging from a satisfactory bus. They might easily save money hanging it from PCI unless PCIe allows multipule devices easily.

As for cards, AFAIK all consumer level machines still use mostly PCI 1x with at most one PCIe for a graphics card.

Servers are a completely different story.

-- Robert

If I recall correctly, you can rent such equipment. Also, most Cat5 wire and such works just fine for gigabit. If you do get the test equipment you can test any old wireing instead of blindly replacing it.

Thomas Schulz wrote in part:

Strongly agreed. Most Cat5 (no e) should also meet Cat5e and run gigabit just fine unless it was very cheap/marginal cable and/or poorly installed (overpulled). Any cables near the length spec (33-') might be marginal.

But testing would save a lot of rewiring.

-- Robert

Still, for some reason not overly rational reason, those sorts of things always remind me of the old Fram oil filter commercials with the mechanic wiping grease off his hands while saying "You can pay me now, or you can pay me later" the implication being spending money on an oil filter now will be lots less than repairs later.

Unless it was really bad to pull new cable, I'd be inclined to pull some of the highest-grade UTP out there today, and perhaps bring some dark fibre along for the ride. But then my day-to-day work-world centers on doing 10G and then 1G more than anything else these days so my perspective may be a triffle off.

rick jones