Infrastructure as a Service
Coming out of VMworld and talking to a lot of cloud providers (CPs) over the past few months, I am seeing many more RFXs and requests for presentations around network designs to support higher capacity compute and storage requirements. I am not referring to CPs selling virtualized servers to the SMB market; I am referring to new cloud providers focused on IaaS. These companies have customers that require large core (20k, 40k, 60k, 100k) clusters for bandwidth intense compute applications. The networks are straightforward in design and with high-end compute and elastic storage requirements targeting pharma/bio-tech, quantitative analysis and research applications. Many have dedicated compute and storage resources for compliance requirements. These CPs are not looking for low latency network switches. They are looking for networks that can dynamically configure high capacity bi-sectional bandwidth on demand to cluster compute and to reconfigure the network to support data replication requirements.
/wrk
* It is all about the network stupid, because it is all about compute. *
** Comments are always welcome in the comments section or in private. **
SIWDT 
Could you be more specific? What is a bi-sectional bandwidth on demand? Are the switches of today not meeting the requirements?
When you build a traditional switched network, it comes with a fixed OSR. Modern compute applications use clusters of servers. I am referring to the ability to have a dynamic layer 1. A underlay network that allows for the physical network topology to change to support modern computing applications.