OpenStack Foundation starts scoping machine learning enhancements
Users want it and NVIDIA's already sniffing around
OPENSTACK SYDNEY The OpenStack Foundation has started to scope what it will take to make its software a suitable platform on which to run machine learning workloads.
“Machine learning this year feels a bit like edge computing last year,” Foundation executive director Jonathan Bryce today told The Register, in that users revealed their work on, or desire for knowledge on, edge computing last year. This year, “I am hearing about a lot of OpenStack users running machine learning. They are asking who else is doing this.”
When the Foundation heard its members asking about edge computing, it started work on features to make the platform good at it, some of which have now arrived.
Bryce said OpenStack hasn't commenced formal machine learning efforts yet, but added “we have had a meeting with NVIDIA, which has started producing drivers that work better in our environment.”
Bryce characterised that meeting and other efforts as “trying to figure out the pieces we should care about”.
He has a inkling of what the Foundation will need to accomplish to tune OpenStack for macine learning. “Users don't want to use VMs, they want a super-abstracted cloud and different types of I/O,” he said. Bryce also knows what at least one of his next steps will be as the Foundation pursues machine learning. Every time we talk to NVIDIA, Intel wants to talk to us too,” he said. ®