Reminds me of my dissertation many years ago; wrote something similar on a network of sun's; something went wrong, process table filled up (fast) on every server (it was only meant to take a small bit from each machine).

I wasn't very popular when every terminal ground to a shuddering halt...

Wonder if they've got some form of RAID happening on the servers? Since the unit has 4 x86's in it, what'll happen if one of them fails? Could be quite a slowdown when the system needs to load in 500/4 GB of data in from some disks somewhere...

Have they thought about a small thing: blackouts.

If RAM loses power, all your data goes poof. NAND Flash memories and such at least have some kind of persistence, RAM doesn't. Ow.

I'm pretty sure they have thought of that. It's pretty much RAM's only weakness.

It's easy to add batteries that keep the data safe for a few hours in case of a power outage. And this thing is a cache, I'm sure it writes up-to-date data continuously to the backing non-volatile storage system.

This could give a big performance boost to any datawarehouse and reporting application.

I know of a couple guys who'd drool if they were told that it's no problem to drill down to posting levels in their accounting system. Today, with nearly 100 mio postings a day, this is not possible.

and by the way: no gives a shit about blackouts in a DW application. nothing's stored anyway.