In telecom use-cases of cloud computing a number of applications have extreme network performance requirements. Largely born in bare-metal, dedicated hardware environments, these applications sometimes find it hard to fit into the virtualized cloud computing model, and operators are confronted with challenges around techologies like virtio, dpdk, sr-iov, the demands for cpu pinning, complexities of NUMA architectures, over-subscription and so on.
This is a proposal for a discussion to include operators, and developers to explore the challenges involved at this important intersection point.
As the telecom applications of openstack continue to gather velocity, having an in-depth discussion of this topic is, I believe, extremely important.
Ideally, I'd have liked to have this as a panel discussion at the OpenStack Summit, I failed to get a talk in on time and I'll definitely do that for Vancouver. But for Sydney, if there's at least a chance to get a forum discussion, it would be a very worthwhile thing.