In case you haven’t seen it, take a look Craig Prescott’s comment at:
http://storagemojo.com/2007/07/25/long-haul-infiniband/

and Craig being quoted in an Obsidian article:
Obsidian Longbow Delivers InfiniBand Storage

Interesting Strategy and some really good numbers in that article.
If you have deep pockets, and need the HPC performance numbers, the long-haul IB is the way to go.

As you mention, there is a “poor-boy” Strategy that can deliver IB cost-justification performance numbers.

Why?

I’ve done some googling around, and there seem to be few articles which compare the two technologies, and those are from 2000 or 2001. I’d love to see a discussion of where how you see ‘the plumbing’ shaking out over the next few years.

Disclaimer: I’m a sysadmin. Until pretty recently, I’ve focused on systems, not SANs. So I’m really coming up to speed on this stuff, but I still have a long way to go.

It seems to me that the deal killer for Infiniband is the distance limitations. In a data center of any size, 17-20m is nothing. The flexibility of 4gbs FC speeds, over 100 and 200m lengths seems to outweigh the very low latency, and higher throughput that Infiniband offers.

Granted, you can do a slower, longer distance version of Infiniband.

To my mind, this seems to doom Infiniband to be a really nice solution for clusters, or for systems that share a rack (or adjacent racks), and have very high IO requirements. However, for a general ‘plumbing’ solution for systems to reach their storage, it doesn’t seem to offer the flexibility needed.