Over at the Go Parallel Blog, John O’Donnell has posted highlights of the year 2014 in parallel computing.
Here are his Top 5:
- Intel Bets Big on Software-Defined Storage: Intel put a bet down on software-defined storage in May by investing in California-based Maxta, which will use the money to develop platform that creates a virtual storage network using storage space attached or assigned to physical servers.
- Highly Anticipated Silicon Photonics at IDF: At IDF in June, Intel gave attendees a download on silicon photonics – the fiberoptic interconnect that has been the star of Intel R&D for a decade and is almost ready to start appearing as part of the high-speed data transfer network for Big-Data clusters and scale-up servers and flat-out supercomputers.
- 2014: The Year SMEs Meet Supercomputers?: In March, Andrew Carr, CEO of Bull Information Systems, predicted 2014 would be the year SMEs discovered the supercomputer, and they sure did. Companies that never thought about supercomputing before were buying or building high-performance clusters and specialty scale-up boxes from Cray and others – mainly to handle Big-Data analytics projects that keep getting more complex because the data get bigger and bigger.
- Ice X Redefines Supercomputer Cool: 3M, SGI and Intel came out with a new immersion-cooling system for supercomputers using a dielectric fluid from 3M, Intel processors and SGI’s new Ice X supercomputer. Called Novec, the cooling system uses one-tenth the space of air-cooled systems, while putting out less noise and using less water and electricity. It would not be the last immersion cooling system to hit the supercomputing world that year.
- HP’s Apollo Supercomputer Aims to Create New Cooling Class: HP launched its Apollo 8000 supercomputer, which is based on Intel Xeon E5-2600 or Xeon Phi processors. It outperforms most datacenters in far less space and saves money, space and power in keeping cool. It uses a water-cooled system that runs heat-absorbent fluid through pipes that run right into the processor nodes to pull as much heat as possible away from the processors while still being hot-swappable without leaking coolant whenever they’re removed from a server rack.