One of the golden rules of network management is to make life as simple as possible for IT staff by standardising on as few types and makes of equipment as is practical. So there may well be many IT managers looking aghast at IBM’s exhortation to prospective users of its BladeCenter H Systems that they should mix and match 10GbE, InfiniBand and Fibre Channel (FC) in the datacentre, creating a "virtual fabric architecture".

This is probably one of the last things many datacentre managers want to hear. Over the past decade or more, many large companies have gradually added all manner and make of equipment to their infrastructure – evolution not revolution is the usual trite marketing phase to describe the process – to the extent that in some cases datacentre managers do not know what is lurking in every rack or cabinet.

It is hard to see 10GbE and InfiniBand being anything other than mutually exclusive in most environments, though there will be exceptions to the rule. However, with FC looking set to dominate the storage area network (SAN) for some time to come, both technologies need to treat FC as a semi-permanent bedfellow, at least until either iSCSI or 10GbE make the impact that some have predicted.

Aside from the management and configuration complexity that mixing multiple technologies causes, there is always the question of whether it all actually works together in the first place.

Even relatively independent interoperability certification, such as that conducted by the Wi-Fi Alliance for wireless LAN kit, never provides absolute certainty that one product can be made to communicate with another at its full potential.

Few datacentre managers will believe that IBM has performed adequate interoperability tests, and the vendor’s claim that it based its analysis on certain configurations demanded by its customers will not offer much reassurance.

That does not mean that BladeCenter H Systems will not deliver the same performance when used with other vendors’ 10GbE modules, InfiniBand components or SAN fabric switches: only that there is no guarantee that configuration and performance problems will not be encountered.

IBM quite rightly points out that it is impossible to test every permutation or combination of kit – it would simply take too long and delay system delivery, probably to the point where the world would have moved on to next-generation technologies.

When all is said and done, you’ll just have to take IBM’s word that at least one firm wanted those precise bits of equipment to work in exactly that way and that is what they tested as a result.