"The SAN term is dead," says John Kelley, McData's CEO. "It's all networks, you're just doing it on different protocols, transparently to the user.

"Fibre Channel moves video as easily as other networks do, the equipment doesn't care how the bits move. We have the ability to move unstructured data too."

Convergence is a big theme with Kelley, and one he's pushing McData to embrace speedily. Last year the company acquired Nishan Systems and has since been busy taking its storage routing technology which enables Fibre Channel and IP networks to talk to each other, and widening its applicability.

This effort has affected McData's finances, pushing operating expenses up and margins down as it spent some $200 million acquiring both Nishan and director developer Sanera, and a stake in chip company Arohi, and then a further $100 million on putting it all together.

Kelley acknowledges too that with everything else that was going on, the company failed to pay enough attention to its biggest customer - EMC. Now though, with cashflow still positive, a healthy $321 million war-chest, and margins on the up, he says that relationship is back on track and he is ready to take the convergence message still further.

He has started out with a number of interesting partnerships. Getting IBM's blue badge on McData hardware is the big one, of course, but he has also sealed deals with QLogic to co-operate on blade switches for blade servers, and for NetApp to rebrand McData's SANavigator management software as its own.

"We love our brand but if we have a chance of carrying the IBM logo, I'd be a dummy to pass it up," he says. "I would love to have a McData-Inside strategy, but I think our customers know who we are anyway. The reality is we're a plumbing company, and storage folks would like to have our stuff just work.

"Value-added resellers and distributors find the McData brand an advantage, so we do have brand equity, but we need to exploit other brands too - it's kind of precarious. 70 percent of what we sell is not McData-branded - we have backed our way into the OEM business."

He adds that 45 of that 70 percent is EMC and the rest will be IBM, and says that he sees a new power bloc forming by partnerships rather than mergers.

"A merger of McData and QLogic isn't in the works, but these things set up symbiotic relationships, and at some point you usually look at deeper relationships.

"Cisco would like to control the network cloud," he continues. "We have plans to help our partners and open our products up so no-one owns the cloud. It should be best-of-breed, and that's not Cisco.

"We are very cognisant of the vendor group that does not want Cisco to dominate. We have access, credibility and prestige in the enterprise and if we can parlay that into partnerships we will have a lot of strong partners. If we roll over, it'll be a very different game for those other companies."

He is very much a fan of IP - he was at Xerox when Ethernet was developed, and at USWest where his team drove that telecoms company's IP infrastructure, choosing Cisco kit as it happens. But he argues that IP isn't ready for enterprise storage networks yet, and that even when it is, it shouldn't really matter what the cable is.

"Fibre Channel and IP won't wipe each other out. I think the networks will blend together - I believe in converged networks," he says, adding that routing technology will enable this, as it does for LANs.

"A routing device enables the exchange of information between SANs, so the customer wants something multivendor-capable, and the brands doesn't matter so much. Think interoperable, like a modem. We've done a lot of implementations of routing for SAN islands in the last quarter. But data may still be stored in the wrong format so that's where you need more layers, of virtualisation for example."

Here he admits that McData is behind: its stake in Arohi has brought technology capable of embedding intelligence within the SAN, but it has not yet brought this to market. Meanwhile, Brocade already has its ex-Rhapsody intelligent switch and Cisco has a virtualisation blade for its MDS9000 platform.

Kelley says McData's intelligent switch will appear next year, and claims that being first to market isn't always best.

"The Brocade and Cisco products still have a way to go as they relate to the user, as would ours," he says. "Early entry here is not as important as cost and interoperability - these things still have too many limitations for practical use. I think both those companies know their products will have to go through several iterations."

And he puts more stress on routing anyhow, saying that next year will see McData put its Eclipse technology into a blade for the Intrepid 10000. CNT already has a similar capability in its UltraNet Multiservice Director, but Kelley argues that McData's iFCP routing protocol is superior to the FCIP bridging technology used by CNT (that is an argument that will run and run, though!).

"It's very comparable to the CNT high-end but the question is when they'll ship," he adds. "We have shipped alpha versions to two OEMs for qualification, with general availability in the fourth quarter. The advantage over CNT [which also sells through IBM] will be the IBM logo on our box."

He says the key thing is combining routing with the i10000's security and logical partitioning capabilities. This will allow users to concentrate multiple heterogenous SAN islands into one network, while still keeping storage segregated where necessary, just as LAN traffic can be restricted to specific segments.

"The biggest growth opportunity is clearly routing," he concludes. "It needs Nortel or Lucent electronics but can then drive 1Gbit/s over 2,700km or 2Gbit/s over 800km. It should begin to solidify enterprise SANs - not just as islands, because now we can connect them securely."