Thursday was the final day of the main conference with the design conference continuing into Friday.
I took the opportunity to browse the large marketplace where vendors make their various pitches. Network Functions Virtualization (NFV) is one of the big buzzwords at this conference. Whilst it will have little impact in the HPC sphere, it promises to revolutionise how telcos operate and to massively reduce the cost in bringing faster communications to market.
The Ceph file system also featured prominently with a number of vendors including OCF’s technology partners SuperMicro and Fujitsu demonstrating their Ceph-optimised hardware. Someone was even demonstrating what they believed to the smallest Ceph cluster in the world on a set of Raspberry Pi computers!
Back at the HPC end of things, the University of Cambridge in combination with Canonical gave an excellent talk on the progress they are making with OpenStack in-house on GPU architecture and as part of ramping up processing power for the Square Kilometre Array. This will output so much data when it comes on stream around 2020 that there are no current computers which will be able to handle it – this is a world of Exabytes of data, never mind Petabytes. I recently gave a talk on OpenStack at the DiRAC conference at Cambridge so was pleased to see that OpenStack is in use there (I was unable to find evidence for this previously) and that great strides have been made already in moving from Proof of Concept to a production environment.
There were more discussions and talks on containers, but I am particularly interested in deploying Manila, the shared file system for OpenStack which will lower the barrier for collaborative research.
It was good to meet some people from the UK who had made it over for the conference and I look forward to doing so again on Friday where developers (hopefully!) agree on development targets for the next release of OpenStack.