In this article from the 5th, http://bit.ly/17GTsst, Marc Cohn says many good things, but seems to feel the bulk of the attention is on the management side of things, and that little attention has been given to Networking:
The NFV Use Cases are intended to drive the Virtualization Requirements, which in turn drive the End-to-End Architecture. NFV encompasses functionality in the compute, storage, and network domains. To date, the majority of the effort has been focused on the software virtualization framework, along with management and orchestration. Consequently, limited attention has been paid to the network.
I’d add that even less attention has been paid to Storage. There are so many things to consider when it comes to that — block vs file, scaleout vs scalein, what protocols, security schemes, etc. The list goes on and on.
Lately, I spend a lot of time thinking about hardware accelerators. The simple reality is the bottleneck is the storage part of the infrastructure, when it comes to virtualization. With that said, we still need the acceleration as well. And much time has been spent in the INF working group to discuss this.
Management and Orchestration (MANO) and SWA (Software Architecture) have their challenges primarily because it’s quite difficult to get multiple vendors to agree on APIs that are normally deemed proprietary. Often, software has to be re-architected to support a vendor-agnostic methodology. And when you consider that NFV is about the WHOLE system, e.g., Compute, Networking, Hypervisor, Application, Storage, Management, Business Logic, etc., that’s quite a lot of cats to try and herd.
Just felt the need to elaborate on his article as the INF working group meets at Brocade for the next couple of days to hash out all of these implications.
One of the reasons I love NFV is it’s where the rubber meets the road.
Wish us luck!