Google Preemptible VMs
Explaining Preemptible VMs: saving 80% on Google compute costs
Google Preemptible VMs are excess compute capacity (Idle On-Demand Severs); across Google’s regions, zones and CPU Platform. Google defines Preemptible VMs as “Affordable, short-lived compute instances suitable for batch jobs and fault-tolerant workloads.”
There’s no bid pricing on Google Preemptible VMs
Even there’s no bid pricing, Google Cloud can reclaim your Preemptible VMs away at any time with a 30 seconds notice. In order to achieve better availability and fault tolerance – it is recommended to diversify your instance types and minimize the risk of losing a bunch of VMs at once.
In addition, Google Cloud VM types are a little more sophisticated than regular Cloud VMs, as VMs are composed of RAM & CPU combined with a CPU Platform and Zone (not to mention GPUs). In order to make the best use of the different permutation of compute that Google can offer in a long-running cluster from preemptible & regular VMs we recommend to use Elastigroup for Google Cloud.
What Else you should know about Preemptible VMs:
1. Preemptible VMs are affordable!
With server costs going at anywhere from 80% off On-Demand prices, and 30-60% off Reserved Instances, Preemptible VMs offer the largest savings across the board.
2. Preemptible VMs can get terminated within 30 seconds
The caveat of PVMs is that Google can terminate these instances with little to no warning (max of 30 seconds). If your workloads require any availability, consistency, or data integrity, managing these workloads on PVMs requires extra configuration and management.