2 ms·
I'm talking about time-sharing. It doesn't mater if it's smaller instances sharing a single GPU instance or many instances sharing many GPU instances. Essential
by mtanski 11y ago
I'm talking about time-sharing. It doesn't mater if it's smaller instances sharing a single GPU instance or many instances sharing many GPU instances. Essentially N:M sharing (with some scheduling).
Since the GPU client is now abstracted from the GPU devices by placing the GPUs across the network. It seams like time-sharing should be next logical step.
- mtweak 11y agoGot it, this is actually already supported. At the very end of the blog post there is a link to create a custom configuration. You can create any N:M configuration, that is any number of clients to servers and therefore the level of performance scaling or GPU pooling. Check it out: https://console.aws.amazon.com/cloudformation/home?region=us-east-1#/stacks/new?stackName=BitfusionCluster&templateURL=https:%2F%2Fs3.amazonaws.com%2Fbitfusionio%2Fcfn%2Fbitfusion-boost-cluster.cfn https://console.aws.amazon.com/cloudformation/home?region=us...
- mtanski 11y agoGreat, thanks for clearing that up.