Announced just before the opening of its HP Discover event in Las Vegas, the company unveiled two new servers designed for high-performance computing workloads as it prepares to take on established supercomputer makers like IBM, Fujitsu and Cray. It’s HP’s first play at grabbing a slice of revenue from what is a highly lucrative market.
HP’s developed some pretty ‘cool’ cooling technology that allows it to pack more computers next to each other, simultaneously reducing the cost of keeping the fierce heat they emit in check. This means companies can achieve massive savings on space and electricity costs with their HP supercomputer installations. This is in contrast to traditional supercomputer builders like IBM or Cray, which are more concerned with custom interconnects that allow scientists to cluster servers together more easily.
The most powerful of the two servers, the Apollo 8000 HPC system, relies on liquid cooling to squeeze a whopping 144 servers into a single rack. The cooling technology is called “dry disconnect”, which involves treated water being piped around the cores, soothing those perspiring chips far more efficiently than any air-based cooling system can.
HP claims its cooling system offers several advantages (you can see a video of it below) compared to systems that submerge servers in special fluids. The main one is that HP’s server blades can be removed without needing to shut the entire system down – something that’s nigh-impossible with a submerged cooling system. In addition, the 8000 is much more efficient on space as it doesn’t need any vats to contain the special liquid for submersion.
HP has paired its cooling system with an HVAC power distribution system and an iCDU Rack to ensure maximum efficiency. The 8000 series uses a technology called “Peregrine”, which was developed during an earlier pilot project HP ran with the National Renewable Energy Lab. That effort helped the lab not only by cutting its electricity bills, but also by using the HPC-heated water to provide heating during the winter.
As well as the Apollo 8000, HP has also launched the Apollo 6000, which is aimed at enterprises wanting to perform tasks like financial service risk analysis and design automation. The Apollo 6000 squeezes 160 less well-appointed servers onto a rack, each with two Intel Xeon E3-1200 processors and two 10GbE FlexibleLOM cards. Future systems will target higher I/O bandwidth applications, HP said.
The Apollo 6000 and 8000 will be available from HP and channel partners from 10 June. This means it’s likely we’ll see it crop up in the world’s Top 500 supercomputer list by mid-2015 – or at least, that’s what HP is hoping for.