Is it possible to abstract the compute or memory of hosts in a data center? I'm thinking of a compute/memory version of a storage array.
Yes, you can virtualize and cluster resources using a hypervisor and all that, but, from my experience, the memory and compute of the physical host are still married together. A VM on host A only has those compute and memory resources available. It could move to host B but then it's limited to those resources.
Also, in a storage array you can replace a failing disk, add a disk, or even add an additional chassis. With the current model you're going to buy an entire additional host if you need additional memory and get unnecessary compute thrown in as well instead of just adding extra sticks of RAM.
So you end up with a physically deconstructed server (memory, compute, and storage) that is then logically reunited in a software-defined sort of manner
#asknostr #it #tech #datacenter #sysadmin #grownostr #nerd #geek #technology
