Dave Bailey

There are more than 10 vendors in the UK selling gigabit, power-over-Ethernet, stackable switches. I know this because I recently contacted them about taking part in a group test. Candidates ranged from reassuringly expensive models from Cisco to cheaper kit from Netgear – all waiting for the day when mainstream companies decide that their end-users finally need gigabit speeds.

There are lots of tools on the market that IT managers can use to determine whether their users need more bandwidth. Any users with throughputs constantly over 50Mbit/s are likely to need more bandwidth pretty soon. If the whole workforce is continually pushing the network infrastructure to its limits, then the chief financial officer may have to bite the bullet and sign the cheque for that gigabit rollout, which could mean buying new cable as well as faster switches. Category 5 cabling that is designed to go to 100Mbit/s may, in certain circumstances, be able to handle Gigabit Ethernet. Otherwise, it’s a case of installing Cat 5e, which adds significantly to the cost of moving to gigabit.

Little wonder then that many network managers are under pressure to make the best of what they’ve got, even if it means some users continually complain to the helpdesk about network speed. (Of course, if these users happen to sit on the board, funds for an upgrade will miraculously materialise.)

But if a business wants to take advantage of the latest in voice-over-IP software or video and web conferencing applications, then upgrading desktop connections to gigabit speeds is more or less unavoidable. But that’s not to say that Gigabit Ethernet networks are without their performance problems. For example, local hard disk read/write speeds are an obvious bottleneck. This isn’t an issue where users write to a hefty storage area network (SAN) of Ultra320 SCSI disks, but that kind of setup is beyond the means of many firms.

Having run a few tests on Gigabit Ethernet throughput, I know that there are other factors that could trip up the unsuspecting IT manager, like network data frame size and the size of the data blocks written to the hard disks.

The natural time to move to Gigabit Ethernet is when old 10/100Mbit/s kit is nearing the end of its working life. But some companies seem to be wilfully blind to the fact that their switches are at death’s door. Sweating assets is one thing, but caning the rocks off something till it drops is not recommended. Some firms appear to be hell-bent on trying to get into the Guinness Book of Records for having the oldest 10Mbit/s switch in town. Maybe they are going to leapfrog gigabit connectivity altogether and go straight to 10Gbit/s?