First up is a variant of the existing build to order program for HPC customers, this one is called Sun HPC Quickstart Services. Sun has long had a config to order program for rack systems, but the HPC Quickstart is a bit more comprehensive. Sun will not only screw the racks together and wire it up, but they will also install the software and tune it a bit.
This may seem trivial, any dolt can screw a 1U server into a rack and connect an ethernet cable. On top of that, if your university buys a supercomputer, you have swarms of grad students to do everything from hand tuning the code to polishing the bolts on the racks for course credits. Life is good.
If you are a corporation wanting to do the same work without an army of slaves for credit, then having a company to do it for you is a big plus. Sun will be a virtual army of slaves for credit, and config it for you. This could be a big boon for businesses wanting to deploy HPC like workloads.
Next up is the hardware, the big news is an updated blade rack. This new blade, the Sun Blade 8000P is an addition to the 8000, not a replacement. What the P does is optimize for density by cutting out all the things the HPC set won't miss much. This mainly amounts to IO capabilities, HPC tends not to use much more than a single connection to the outside world.
The 8000P has about 50% greater density than the 8000, but uses the same blades. The midplane and chassis are all different, lacking a lot of frilly extras. They goal is to separate out the compute functions from the IO and allow for long term expandability. They added PCIe, allowed for PCIe2 in the future, and added room for more DIMMs.
IO was reduced and aggregated while overall power was reduced. Some of this is from the transition from Socket 940 to Socket F and some of it is from reengineering. It also showcases the neatest feature, you can mix and match Socket 940 and F Opterons, basically some Rev E blades, some Rev F, and they play nice together.
There are a lot of partnerships to go along with the hardware and services. First is NEC who worked with Sun on the TSUBAME supercomputer at Tokyo Tech. The knowledge and deployment expertise picked up there will also be available to HPC customers.
What good is all this hardware going to do you if you don't have good tools? That one is also easily solved by a new version of Sun Studio and Cluster Tools. With the speed of advances in threading, clustering and HPC optimizations in general, these new compilers are worth a look no matter what you are doing.
In general Sun it taking what it learned on the high end parts and bringing it from academia to business. The firm is rolling out all parts of the chain from the roll-out itself to coding, all in the hopes of making your life easier and more efficient. Oh yeah, they want to make money too, but if they do the easy and efficient thing, they probably will make some of that. ยต