Michael Arrington had some questions about yesterday's announcement. I've attempted to provide some answers below.
1. How do third parties open their own facilities/data centers? What if Stanford wants to open a facility?
Third parties are welcome to join the test bed by contacting the founding corporate sponsors (HP, Intel, and Yahoo)
2. Pricing: How will resources be allocated to people building on the platform?
This is a research test bed and resources will be allocated based on research activities. The intent of the test bed is to support research into large-scale systems. To avoid dividing the resources up into many small slivers, we will bias the allocation policy to support short, large-scale experiments, rather than smaller, long-lived ones. We hope to construct several services that will be used by many of the participants; to ensure that we don't consume all the resources for such services, they will be treated just like other users of the system, requiring them to acquire resources from the common pool. In this way, we hope to be able to manage allocation, and flex resource needs as a function of popularity and load - just like other cloud computing services.
3. Where do people go for information on APIs and other tools needed to access the platform?
We have a web site under development that will serve as a resource for information about the test bed. The website will have details about APIs and tools shared by all members, news, resources, and a discussion forum.
4. Yahoo kept referring to M45 as a version of this already deployed. Is that their contribution or are they building out a new facility?
The M45 will be used as part of the test bed.
5. Overall this is super squishy, and appears to be more of a hype release than anything. More details are needed. A lot more.
The test bed will have six centers of excellence that are expected to be operational by the end of the year. On the technical side, the entire stack, from applications to systems software to lower layers, is expected to be available for investigation. We expect to support APIs that range from enabling users to reserve, reset, reboot, power up, and power down VLAN-isolated machines, to higher level services like virtual machines on demand, virtual resource services, and the ability to run Hadoop and/or Pig jobs. Given that this is just the announcement of the test bed, more details are definitely on the way. With major commitment from companies, universities, and government agencies, we are already in the procurement phase and expect the facilities to be up and running later this year. We’ll make sure to send you updates as we reach other milestones.