It's now a year or two since Dell made the move into blade servers, so has it succeeded – or at least started to succeed? And if so, does it matter to anyone apart from those at Dell?
Take blade servers. A Dell server product manager I spoke recently reckoned that, although even small to medium-sized businesses are buying them, most buyers are big companies. Smaller companies, such as media and graphics production houses, do use them to get high-density raw compute power, but that kind of user is few and far between. If you're running out of space to put your servers and you have the space, infrastructure and financial wherewithal to cool a rack or three of blade servers, along with the associated power and storage infrastructure, then you're likely to be a fairly hefty organisation.
The issue with blades is that they're sold on the razor blade business model. Buy a chassis from a blade vendor and you're locked into that vendor's product strategy and price list. You can't swap severs across different vendors' chassis. In fact, in some cases you can't even swap servers across chassis from the same vendor. There's no sign of inter-vendor interoperability arriving in the near to mid-term future, although Ed English, Dell's European server product manager, would like it to. Of course he would: it would give him access to the market leaders' customers.
Rivals HP and IBM – for they are the head-to-head market leaders – would surely not welcome this move.
Funnily enough, might some customers also not welcome it? On the face of it, it sounds like second sourcing and price competition would be a good thing in the blade market, where currently there is none – on the server front anyway – although there's plenty of competition inside the chassis for the I/O, with Brocade, Cisco and a host (no pun intended) of others competing by producing IBM and HP-compatible components. They're making kit to fit into Dell boxes too.
But would a common interface work for servers? More than one company using blades has pointed out that, when clashes or glitches occur, it would be tough to figure out where to point the finger. Of course this bolsters the argument for a single-source, end-to-end product acquisition strategy, something enterprises are already comfortable with – as long as there's been a fierce (but civilised, please!) scrap for the contract in the first place.
But since enterprises are now moving in the direction of outsourcing, they're increasingly likely to outsource IT maintenance if not infrastructure to a third party integrator and services company. It's then up the service provider to sort out interoperability problems, so a mixed server-chassis system need not be a problem. So companies using mixed architectures as a rationale for not wanting competition in the blade server space could be swimming upstream.
As a result, I think that competition for the space at the front of the blade chassis is overdue – but despite Dell's fervent desire for it to happen, it remains unlikely. There's still too much in it for the vendors.