When using public cloud computing, clients are often under the impression that there is unlimited capacity on the major vendor platforms such as AWS or Azure, available whenever you need it.
However, due to the popularity of many popular “instances”, capacity is not actually unlimited, there is a finite number available at any time in any particular public cloud data center.
For this reason, from time time to time you may experience capacity issues when there is more demand than capacity. For example, on AWS this is the error you will receive when trying to spin up instances when there is no capacity available:
If you need the on-demand access “right now” to get an answer back, this can be a bit disconcerting. AWS will start killing spot instances to free up capacity, and then you will be able to get the instances up to an hour later.
Using a vendor like TotalCAE we can avoid these scenarios by spinning up our public clouds in multiple public cloud data-center locations if customer chooses. or even completely different public cloud vendors including Azure, thus avoiding any delay in running your on-demand jobs if time is critical.
Alternatively, you can also not “give up” the nodes. If you are not using spot instances, you are guaranteed the nodes will not be taken from you, but you are doing pay as you go. You can also pre-pay for reserved instances, but then that defeats the purpose of pay as you go as you have basically bought the nodes.