The following text is copyright 1996 by Network World, permission is hearby given for reproduction, as long as attribution is given and this notice is included.

Selling Virtual Spaghetti

The other day I happened across Charles Feltman's article in the July 1996 issue of Business Communications Review entitled A Reality Check on Virtual LANs . It is quite a good article but I think Charles missed one issue that makes the future usefulness of VLANs hard to predict.

The article explores some of the reasons why the VLAN concept is seen as useful and some of the reasons why the concept does not seem to have taken off.

For the jargon deprived, in this context I'm considering a Local Area Network (LAN) to be the equivalent of an IP subnet or an AppleTalk or Novell network.- That is, the network or pieces of network connected by bridges or switches over which a packet sent to the broadcast address will be forwarded. LANs are defined by the physical topology of where the network wires run. A Virtual LAN (VLAN) is a LAN that is not constrained by the pattern of physical wires. VLANs are implemented by connecting the network wires together with switches which can logically group nodes together even if they are on separate wires.

The primary advantage cited for VLANs is the ability to deal with physical and logical personnel movement without recabling the networks. For example, Fred moves down the hall to an office with a window (in lieu of a raise) but Fred should still be considered part of the same group network-wise. It is claimed that this ability to deal with moves, adds and changes with far less effort could significantly reduce networking support costs in some of today's companies with all too active org charts.

Why might it be important to continue to have Fred be part of the same network group? There could be a number of reasons, from separation of users for security reasons to not wanting to have to go reconfigure the IP address in Fred's PC. But the most common reason cited is to keep the users of a server grouped together so their traffic is isolated from the rest of the network.

The use of DHCP can, with reasonable ease, eliminate the requirement to reconfigure individual machines. I'm not so sure that it is easy to group users into security-related groupings so that leaves the user's affinity to a particular server as the major rational for VLANs.

But what about the trend toward centralized servers in many corporations? Centralized in two senses of the word, putting the servers in a central location and putting more server functions on the same platform. I don't see the advantages of VLANs over physical LANs if one can not logically associate groups of users with individual servers. If most users use the same servers then the division of the users into groups will not have the desired effect of separating the network traffic to reduce the network load.

It seems to me that selling virtual spaghetti (which is what VLANs are) is going to get harder as time goes on if the trend of concentrating points of failure continues.

An unrelated observation about the TV coverage of the recent Olympic games, it sure would be better if NBC paid their commentators by the hour instead of by the word.

disclaimer: You can't use centralize and Harvard in the same sentence so the above must be my ideas.