How big can we scale?

From the outset of Vyatta we've been focused on the small-to-medium enterprise (SME) market for our initial products. I've written here about how we don't see our focus on the Small Office/Home Office (SOHO) market nor on the Internet core. We just don't see how the commodity hardware that we're testing could scale down to the SOHO market or scale up to the needs of the new monoliths like AT&T and Verizon.

Yet, recently, I've been hearing about folks that are building large scale platforms with fiber-based switch fabrics and lots of compute power that run Linux as their core operating system. I'm specifically talking about people like Rackable Systems, Verari, and Liquid Computing. Today, most of these systems are focused on being utilized for high performance computing (HPC) or corporate grids, but I don't think that it is a far leap to think about these platforms being useful as a distributed routing/firewall platform.

I can imagine a hardware platform with non-blocking fiber switch fabric that scales to dozens or even hundreds of compute blades working in unison. The compute blades could be proprietary hardware or even commodity off the shelf, depending on the reliability and performance needs. Connect that platform with a few 10GigE ports to a network and I think you may have the next generation carrier routing system. Vyatta running on Linux in that environment should be quite scalable. And, just for fun, let's run Xen and virtualize the whole thing! Thoughts?