For years, we’ve been hearing how this year (for all values of this year) is the year 10GbE takes off. I’ve commented on this a number of times, from the context of 10GbE breaking out in clusters, 10GbE killing off infiniband, etc.
Looking back, these comments extend 6+ years into the past. The point I have always argued as being the most important, has been cost per port.
Well, the technical press noted this today.
Mass adoption requires very reasonable and dropping costs. Its not quite there yet as they note:
I think the pricing per port is a little off, I’ve seen good switches go for under $400 for 48 ports. A 64 port Mellanox switch can be had for about $10k USD, or about $156/port. Its still high, but a 36 port FDR Infiniband switch comes in at about $6600, or $183/port, so its reasonable … with the caveat that the Infiniband switch is half an order of magnitude faster in the best case. Really we should be comparing 40GbE ethernet to QDR/FDR IB, and there, IB wins on pricing, hands down.
This said, this factor of 20x above market pricing for 1GbE connections probably won’t survive long. I think we could see it getting down to 10x and below over the next year.
But I don’t think 10GbE will kill off Infiniband, or have a massive break out year until the pricing per port starts getting within a factor of 2 or 3 of reasonable 1GbE costs per port. I don’t see this happening soon.
So I do agree with the article in that it likely won’t happen this year. And it really is all driven by cost. Everything is.
This said, I love that the stack is a single driver in most cases, and it just works (its ethernet after all). Infiniband can be (and often is) maddening. I’d prefer easy.