wolfkillj (1/31/2013)
This sounds like a scaled-up version of the blade server concept. Old model - when you need extra computing power in the accounting system servers for end-of-month processes (the scenario commonly used to explain and sell the blade concept), pull some blades from a system with less load and add some to the accounting system. New model - when you need extra computing power in the data center, truck in a couple of containers of server racks and wire them in.I also wonder whether there could be some DR/business continuity applications. A warm site could be just a big room with the storage infrastructure, the minimum number of servers needed to manage the data replication, and the necessary power and connectivity infrastructure, plus a contract with a vendor to deliver x containers containing y servers provisioned with z configuration within n hours to bring the warm site online to handle the production load.
Interesting stuff.
We used to have an IBM P690 like this, basically a large computer that could be partitioned with a native hypervisor. In 2000 time frame, we had 32 CPUS and like 96GB of RAM. We had this partitioned with AIX into 3-4 VMs that people used. We could shut down some of the text environments for EOM or EOY processing and hot add CPUs and RAM to the production machine.
These days I'd think we should have more cloud oriented software we can add/remove machines from.