That's the first time I've head that !
Wll, these are the workhorses of scientific applications, and represents one of the biggest collaborations for e-Infrastructure in the continent. There's a very solid and mature stack of middleware which they are built with, and most of the configuration is automated. Basically what researchers get access to is a huge pool of HPC resources and distributed data storage. This is a production service, which is community-agnostic, so anyone should be able to use it, and contribute resources to it. Also, support by the RRENs of the ROC is one of the most important aspects of sustainability that came out of the CHAIN-REDS project.
The difference between this "grid" infrastructure and something more cloudy is that it provides specific services, while cloud services are far more diverse. While this makes for much more flexibility, it also means much of the overhead for managing and operating these services lies with the user community.
Cloud services do exist within the ROC. Some of the sites are now providing FedCloud compatible sites, which accept VM requests via OCCI. This is as far as I know the only use of these cloud services - opportunistic computing with pre-configured VMs from a registry (mostly the AppDB). We also do development on an OpenNebula cluster in Pretoria, and there are a few Open Stack sites in and around the region.
You mention SWITCHEngines, which is if I understand compute on demand, which is a very widely-used concept. The most recent iteration of this kind of service, I think is EC 3, and there are many more.... none of this stuff is really different from what we have been doing for almost 10 years in the distributed computing world.
The thing that is different, and which needs to be supported, is the scale at which we provide resources. It is of no use to anyone to have lots of little pools of resources around the continent, independent of each other. The reason that the ROC exists is to act as a focal point for pooling both hardware and human resources, which then interface as a single entity to international infrastructures like EGI. The only reason we can say to a researcher in Malawi, or Benin, or anywhere, that if they want to access resources over the network then they can, is because we have all of these integrative services allow that. Else everyone would be relying on the little bits of kit that they can get their hands on locally.
So, yes if researchers want cloudy services, then by all means ask for that and we will find a site that can provide it. They will be using the same authentication, accounting and support services that the rest of the users rely on.
if people want :
then we need to know about it. However, we do need to differentiate between commodity and specialised services. If a community needs a website, there's no reason that this has to be run on a site in the ROC - you can get one from anywhere. However, when it comes to running scientific workloads, storing scientific data and so on, there is far more imperative to do this with resources in the ROC, since we can provide a more specific support and service. For example, bringing research applications or workflows into user environments, or putting communities in touch with each other in Virtual Organisations - these are the kinds of things which set these infrastructures apart from commercial offerings.
Cloud or grid or anything - it's not going to help anyone really if there's no collaboration at the technical level. This is why we have a ROC.