A cloud myth is that cloud computing offers everyone unlimited capacity whenever you want it, but the cloud vendors sometimes reach capacity limits on many popular HPC instance types for CAE. The meteoric rise in the popularity of the cloud creates an occasional queue for on-demand hardware resources in the cloud, as the cloud is fundamentally just a remote data center with computers in it, and there is a finite number of the specialized computers used for CAE.
TotalCAE 2023.3 Platform release enhanced feedback for cloud clients to alert them that a job can’t immediately schedule on-demand due to their AWS or Azure resource constraints and update you as it keeps the job queued while it waits for the cloud infrastructure to release more hardware resources.
Tracking Out of Capacity Events
The major cloud vendors do not publish the amount of any particular instance in a region or how many are free at any specific point. TotalCAE works with clients and their cloud reps to ensure we are provisioning into the freest regions to which the client can deploy.
To assist clients in tracking any temporary lack of capacity for their preferred instance, TotalCAE graphs capacity denials as part of our Analytics to inform users that they may want to use a different but similar instance. For example, perhaps an older Azure HBv2 might have more availability recently instead of Azure HBv3. An example of the kind of Analytics report is below by region, AZ, and instance type:
TotalCAE Analyzer tracks out of capacity by region, AZ, and instance type.
A Note on Hardware Instance Quotas
Note that by default, the cloud vendors restrict the type and quantity of hardware you can provision with quotas, you have to request access to be able to provision most CAE instance types for review by their cloud capacity team. One major cloud vendor denies quota increases for HPC instances without invoice billing ( no credit cards allowed to be used) though this isn’t an issue for most enterprise client accounts.
In most cases, increasing quotas is simply opening a ticket with your cloud vendor of choice ( though quota increases may be denied, and may take an appeal). We work with clients to assist them to get their quotas increased in their AWS or Azure subscriptions for the types of instances used in HPC. However, note that a quota increase does not guarantee that capacity it just enables the ability to request the capacity up to your quota limit.
We expect the cloud vendors will catch up with the demand for popular instances, but in the interim, TotalCAE is making it easy for cloud clients to monitor and work with the occasional on-demand cloud capacity constraints. If you are interested in having TotalCAE to help with your on-prem HPC cluster or on-demand cloud in your cloud subscription reach out to us at firstname.lastname@example.org