They have lots of cores and are very NUMA.
RHEL 6 = Red Hat Enterprise Linux NUMA = Non-Uniform Memory Access
NUMA is 90's tech from Sequent.
The big HP NUMA boxes are Superdome and Superdome 2, and those have global memory access, though access to the local memory is faster than the remote memory; the requisite non-uniform memory access speeds.
Last I checked, the current Superdome and Superdome 2 series maxed out around two and four terabytes of physical memory.
If the current four terabyte maximum memory configurations and possibly the architectural 50-bit addressing aren't enough and you're willing and able to buy that much physical memory, then HP would probably interested in chatting with you.
No RHEL on Itanium, though.
From: http://www.theregister.co.uk/2010/04/21/redhat_rhel6_beta_1/
This sounds more like Reddit's problem where some architectural simplifications might net a giant win versus piling yet more gunk on top (Reddit is still perceptibly doing random IO for every comment in a thread during page load, or perhaps some insanely slow sorting, I have NFC how they haven't fixed this yet).
They ended up splitting the memory between a few machines as it became obvious that the applications being used (Adobe CS stuff) wasn't going to use all that RAM in their use case, and other machines needed it more.
The details will be interesting, but that description sounds like a headache waiting to happen.
The CPU is organized that way and it works very well and that caching scheme is implemented directly on the metal. Any decent software engineer should be able to design a working implementation with an architecture that is clean.
That's more RAM than our entire system had.
16GB DDR3 is "only" $100, that would make non-ECC $5400
ah okay, ECC is $175, so 175 * 3 * 18 = $9450
This is what 864GB of RAM looks like now
Content fetched 0 seconds ago