* More reader responses
Last time, I shared some feedback from readers on the issue of Gigabit Ethernet to the desktop and future-proofing. This time, I’ve got a few more responses.
One reader sees a parallel between what’s happening now and what happened with the last round of upgrades:
“Back in 1999 our company installed a new network with a dual Gigabit Ethernet backbone and 10/100Mbps to the desktop. Most of our wiring back then did not even support 100Mbps to the desktop, but we installed it anyway because the price was attractive enough.
“Today, we are in the process of upgrading that aging infrastructure out of necessity, and the price for 10/100/1000Mbps to the desktop is attractive enough to be a requirement even though we don’t necessarily need the higher speed. We will have to keep the network a minimum of five years before the next upgrade, so bang for the buck is one factor that we consider, albeit not the only one.
“A great benefit that we are seeing is that some of the technology that we would have had to purchase separately is now being implemented directly in the networking gear. So for us, intelligence weighs just as much as feeds and speeds.”
I also received a message from Geoff Thompson at Nortel, who identifies himself as former chair of the IEEE 802.3 and has an interesting take on the role of Ethernet:
“The job of Ethernet is to provide computer-to-computer communication and to be fast enough to stay out of the way. Thus, the appropriate question with respect to Gigabit is whether or not Fast Ethernet as a desktop connection ever slows down a transfer. This applies to burst applications (i.e. file transfer time) as well as streaming applications. Streaming applications are seldom the applications hitting the limits (with the exception of over-the-net backup) of the end link. If the end link runs at saturation, then spend the small amount of money that it takes to upgrade. Let some other element in the system (e.g user software bandwidth) be what constrains the speed.
“At some point this will all end for other reasons. We are already in the neighborhood of the limit of humans to assimilate information (I benchmark this at ~ 2.5 Gig based on the highest bit rate of uncompressed HDTV). After that, it is all about how long it takes you to download your own copy of the full set of Harry Potter movies. Again, I claim, let some other element in the system be what constrains the speed.”
Another reader looks at the potentially substantial cabling costs:
“While future-proofing may be a good idea, one must always consider all associated costs. For example, if it is decided to actually run Gigabit Ethernet to the desktop, will you have to replace the horizontal cabling? There is a large installed base of Cat 5 cabling that will support 100Mbps and may support gigabit. If your Cat 5 cabling fails to meet the requirements for Cat 5e and must be replaced to support gigabit, or if you decide to replace it as part of your effort to future-proof your network, these costs must also be considered. For large projects, I estimate the cost of Cat 5e or Cat 6 cabling to be on the order of $4 per square foot of space served.
“The price tag may go up substantially should you decide to provide cabling that will support 10 gigabits and install new augmented Cat 6 cabling. Aside from the increased cost of the cabling and its installation, the outside diameter of this cable is substantially larger than that of Cat 5e or Cat 6 cabling. It may well be that your pathway system (conduit or cable tray) has insufficient capacity to carry this larger cable. Consequently, you must include any costs of augmenting or replacing your pathways in determining the overall cost.”
Yet another reader points out that controlling QoS isn’t free:
“As an operating system component builder and system programmer early in my career I spent lots of time diddling with priority mechanisms and tuning same, and life is too short. As with any reservation or prioritization scheme, QoS is great when you are in a sweet-spot capacity-wise, but it’s not magic. It doesn’t increase effective capacity beyond a certain extent, and it can take work, which costs money. At some point you need capacity to run economically.
“Keep the ‘High-Speed LAN’ emphasis – build it and they will come
“I think that your newsletter fills a necessary role, and discusses a lot of the specific issues encountered in trying to anticipate the capacity curve.”
I couldn’t agree more. Many thanks to all who wrote in.
