Spot VMs, previously known as preemptible VMs, are spare compute capacity offered at a significant discount. When we launched Spot VMs last year, we removed the 24 hour limitation so Spot VMs can now run indefinitely. We continue to offer the industry’s most predictable, simple, and competitive Spot pricing – with pricing updates implemented up to once a month.
As an example of bringing savings to you, today, we are excited to share that we have lowered prices on standard and custom VMs and memory in 37 resource pools (where a resource pool refers to a specific VM resource in a given region).
While you can find prices on the VM instance pricing page, and Cloud Billing Catalog API, a summary of the price changes can also be found in the table below. As you can see, we are offering additional discounts off On Demand – as much as 11% on T2D in certain regions! Because T2D VMs can provide over 40% higher price-performance when compared to other leading clouds, you can do even more with less. To stay updated on Spot pricing, you can sign up for Spot VMs pricing announcements..
The below prices apply to all VM types, including custom VMs, high CPU, and high memory variations. While Spot VMs are equivalent in performance to on demand VMs, they are not guaranteed capacity, so we recommend being VM or region flexible. See below for more tips and resources.
Note: though Spot discounts apply to GPUs and local SSDs, today we are not updating those prices.
Customers across the globe have been transforming and innovating faster by using Spot VMs.
“Media.net is a global company with dynamic resource requirements. [We] chose Spot VMs after exploring various options to support spiky workloads, as they provided Media.net with both deep discounts and simple, predictable pricing.” — Amit Bhawani, Sr VP of Engineering, Media.net
“Businesses rely on LiveRamp to process petabytes of file and streaming data safely, securely and predictably to generate valuable business outcomes. We run tens of Dataproc clusters with thousands of nodes and tens of thousands of cores at a given point of time. We extensively use Google Cloud Spot to reduce our cost and to meet the variability and spikes in customer demand.” — Mithun Bondugula, Head of Big Data Infrastructure, LiveRamp
Here are a few best practices to consider when choosing your Spot VM:
If you are using preemptible VMs today, migrate to Spot by updating your API calls to provisioning model = SPOT to ensure your VMs run can run beyond 24 hours (documentation)
Take advantage of our lowest priced SKUs, in particular those prices dropped today.
Select at least two resource pools to minimize interruption, and use regional MIGs with zone = ‘ANY” to minimize interruption (documentation)
When you select this distribution type, the group picks zones for creating VM instances to fulfill the requested number of VMs within present resource constraints and to maximize utilization of unused zonal reservations.
Use BulkCreate API, which allows you to create multiple Spot VMs with a single request. Like regional MIGs the zone with available Spot capacity is automatically selected.
Deploy Spot VMs using Batch. Batch is a free native Google Cloud service that handles the scheduling and execution of your workload while provisioning VM instances, and automatically use the -ANY distribution to reduce interruption
For more resources and best practices, please check out our latest blog posts below.
Cloud BlogRead More