Combine Multiple AWS Instances into a 16-GPU Monster Machine • Hacker News Board

Re: Combine Multiple AWS Instances into a 16-GPU Monster Machine

#41

Post by joombaga » Fri, Apr 01, 2016, 4:55 AM UTC

Johnny555 wrote: post #35

eximius wrote: post #27

Earlier quoted context omitted.

That sounds like a nightmare. You'd have zero uptime guarantees!

Does Amazon have any instance uptime guarantees?

Yes, >99.95% Monthly Uptime Percentage.

"“Monthly Uptime Percentage” is calculated by subtracting from 100% the percentage of minutes during the month in which Amazon EC2 or Amazon EBS, as applicable, was in the state of “Region Unavailable.” Monthly Uptime Percentage measurements exclude downtime resulting directly or indirectly from any Amazon EC2 SLA Exclusion (defined below)."

https://aws.amazon.com/ec2/sla/

Re: Combine Multiple AWS Instances into a 16-GPU Monster Machine

#42

Post by homero » Fri, Apr 01, 2016, 6:01 AM UTC

You can do this 10x cheaper at home

Re: Combine Multiple AWS Instances into a 16-GPU Monster Machine

#43

Post by 3legcat » Fri, Apr 01, 2016, 9:10 AM UTC

manav wrote: post #14

I've found amazon GPU instances to be really expensive (even the spot prices have been high recently), especially if you need it for longer runs for deep learning. The other issue is that the additional layers of virtualization create bandwidth overhead issues. I'd like to see something in the cloud thats bare-metal / full access to GPUs (Maybe a good idea to start one). For scaling higher with a very large number of…

Softlayer has something like it:

http://www.softlayer.com/gpu%20

Re: Combine Multiple AWS Instances into a 16-GPU Monster Machine

#44

Post by man5quid » Fri, Apr 01, 2016, 1:53 PM UTC

Congratulations it's awesome to see the AMIs published.

Re: Combine Multiple AWS Instances into a 16-GPU Monster Machine

#45

Post by mtanski » Fri, Apr 01, 2016, 3:07 PM UTC

mtweak wrote: post #40

mtanski wrote: post #36

Earlier quoted context omitted.

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.

Got 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...

Great, thanks for clearing that up.

Re: Combine Multiple AWS Instances into a 16-GPU Monster Machine

#46

Post by flamethrower » Fri, Apr 01, 2016, 5:14 PM UTC

When will Bitfusion be available for Google and Microsoft?

Re: Combine Multiple AWS Instances into a 16-GPU Monster Machine

#47

Post by Johnny555 » Fri, Apr 01, 2016, 7:46 PM UTC

joombaga wrote: post #41

Johnny555 wrote: post #35

Earlier quoted context omitted.

Does Amazon have any instance uptime guarantees?

Yes, >99.95% Monthly Uptime Percentage. "“Monthly Uptime Percentage” is calculated by subtracting from 100% the percentage of minutes during the month in which Amazon EC2 or Amazon EBS, as applicable, was in the state of “Region Unavailable.” Monthly Uptime Percentage measurements exclude downtime resulting directly or indirectly from any Amazon EC2 SLA Exclusion (defined below)." https://aws.amazon.com/ec2/sla/

That's a regional outage, they provide no SLA for individual instances:

Amazon EC2 SLA Exclusions... (v) that result from failures of individual instances or volumes not attributable to Region Unavailability

Presumably if you're using a cloud hosted in people's basements, if someone's basement server dies, you'd just pick one from someone else's basement, so this model could provide better availability than AWS.