Need to test network

jm,
Sounds like fun! I think you'll need some sort of "sniffer" device to get all of the information you want. We have used a Network Associates' dolch - a dedicated sniffing box, but it doesn't perform the cable testing you mentioned. It was also pretty pricey ($17K). There's a new product out by Fluke called the "Optiview". You can find more about it at http://www.flukenetworks.com/lan/optiview/index.htm. It looks like a sweet box and does perform cable testing along with packet capture and decode, plus analysis. Some of our techs have been salivating over it for some time. We haven't been able to get ahold of a demo unit yet, but I believe its price is around $13K-$15K. We recently replaced our Network Associates Distributed Sniffer with a bundled software package from Wildpackets.com. It had their Etherpeek, Netsense, Pro-Convert and Internet Tools software along with 2 years of maintenance for around $2k/copy. Etherpeek does packet capture and decode; Netsense does analysis; Pro-Convert converts packet traces by other vendors to etherpeek format and the Internet Tools are a bunch of handy utilities for network troubleshooting...traceroute, ping, etc...

Concerning the rest of the justification you have to provide, have you considered deluging him with white papers from various manufacturers of switch gear? They usually prepare comparisons (favorable to their brand of course) with other types and models in the same equipment class. You might also want to search out www.tolly.com. This is a reputable independent testing firm that puts manufacturer's equipment to the test.
(more to follow)

(follow-up to the first response by "Transient". BTW, who decides the cut-off time? This is the 2nd or 3rd time I've been logged out while preparing a response to a question and it gets attributed to "transient". Really ticks me off!)

I would recommend you spend some time on Cisco's website. They have almost too much information, but they also have design guidelines for networks. Basically, any ethernet segment will experience performance problems once it hits 40% utilization. Token Ring and Serial lines are good until 70%. They have proven these figures through extensive testing.

One of the problems your boss has set before you is that field test. The hub has to be under a good load before you'll see much of a performance difference. Most switches nowadays start with 16 port capacity and go up from there. Can you get 16 machines and the people to man them for a field test? If you're like most school districts, you're lucky to have equipment, let alone spares. Try toexplain the difficulty in providing a proper test environment on site.

Here's another idea...make up a short list of vendors and contact them to see if they can provide you with information from installations they have completed at other companies. More than likely, they had to prove their case to someone else before they got the go ahead.

We revamped our old hub network two years ago and have never looked back. Best wishes,
ck

While a comparison between a 100M switch and a 10M hub SHOULD be a no-brainer, I can't tell you how often I've been unable to demonstrate the advantage because of NIC cards that won't auto-negotiate, a switch with an inadequate backplane for the number of ports, etc.
I have a couple of suggestions. First, when I've been doing work on behalf of a governmental or non-profit organization, I've always been able to find a sales rep willing to loan me a piece of equipment to test. Design some specific, real-world tests of things that users do that stress the system (printing a 50meg bitmap, f'rinstance...), establish a baseline, swap the hub for a switch and repeat the test.

I would also try to narrow down the bottlenecks by physical location based on user complaints. If they are complaining that Internet access is slow, I would measure ping times and maybe do some FTP transfers from a known stable site using a client like WS_FTPLE that reports transfer speeds and compare differentsegments.
If there are complaints that server access is slow, I would use Performance Monitor (assuming NT servers) to measure server performance and make sure the bottleneck is not at the server. I might even set up some tests of transferring a large file from one workstation to another to take the server entirely out of the equation.
There is no magic bullet. I have access to the best test equipment and software that money can buy but I find that most network problems are identified and diagnosed by user-level testing rather than by packet capture or other expensive technologies.
However, if you have to have graphs, charts and numbers, I would recommend a free product like MRTG which will provide 98% of the information that a $15,000 piece of equipment would.

paul