Powerconnect 2016

There are multiple reasons you would see a performance loss on a small network when replacing a hub with a switch. Some of the most common are:

1. Our switches use the store-and-forward switching method. This means that they copy every frame into the buffers, run a CRC to check for errors, then either drop or forward the frame. A hub is simply a physical layer device and is basically just an extension of the cable. It has no buffers and simply floods all traffic. If you have a large amount of errors on your network, the switch may be dropping those frames which forces the clients/servers to re-transmit them.
2. Since hubs make no intelligent decisions, they are not sensitive to sub-standard cabling. Switches on the other hand are very sensitive to this. Sub-standard cabling can cause errors which would cause the switch to drop the frames.

I would suggest updating the drivers for all of your NICs, and try using store-bought pre-fabricated cat5 (or higher) cabling.

The cables are store bought CAT5 's.  All NIC drivers are current.

Any other ideas? 

Thanks

McBill

You could try analyzing the traffic to the server to look for excessive TCP retransmits from a particular source. To do this you would need to put the hub between the server and the switch and connect a client to the hub with a packet "sniffer" application running.

A switch is a far more efficient device than a hub due to intelligent forwarding decisions and scalability, but the switching intelligence may increase the latency of traffic forwarding compared to a hub. However, this should not affect any real world applications in terms of performance.

Message Edited by DELL-GregG on 03-04-2004 01:45 PM

Try using Cat 5e and not cat 5 cables.  I don't think cat 5 is certified for Gig Ethernet.