Platform9 is moving to give organizations that rely on its managed service to deploy Kubernetes more control over costs.
The Platform9 Managed Kubernetes service now allows organizations to define what percentage of their total number of nodes in a Kubernetes cluster they want to deploy as Spot instances on Amazon EC2. Spot Instances are available on-demand at different price points based on overall availability of compute resources on the AWS cloud at any point in time.
Platform9 CEO Sirish Raghuram says an Arbitrage capability added to the Platform9 Managed Kubernetes service can now continuously monitor Spot and standard EC2 instances to determine where the most cost-effective location is for a Kubernetes cluster based on the service level agreements (SLAs) established. In addition, organizations can employ that capability to maintain specific levels of fault tolerance and performance, he says.
Arbitrage also integrates with Fission, the open source serverless computing framework for executing functions using spot instances on EC2, Raghuram notes.
The Platform9 managed service differs from managed services provided by cloud service providers in that Platform9 provides customers with multiple options to deploy Kubernetes spanning different public clouds and on-premises environments. Raghuram says Platform9 plans to extend the Arbitrage capability out to other public clouds, which then will make it possible for organizations to essentially conduct a reverse auction whenever they decide to deploy a Kubernetes cluster. All things being equal in terms of performance, organizations will always be able to guarantee the lowest price possible by driving cloud service providers to bid on running clusters, he says.
Platform9 Managed Kubernetes provides the console required to keep track of all those deployments to ensure various SLAs are met, Raghuram adds.
Most organizations have not been able to effectively drive down prices for core cloud services very much. Cloud service providers have been dropping prices mainly for ancillary services, while making available spot pricing to make excess compute capacity available whenever necessary. But because Kubernetes clusters can run anywhere, it becomes much easier to move workloads between different cloud service providers. Most organizations are not likely to shift workloads between cloud service providers once they are deployed, but they can make cloud service providers bid to run workloads before they are deployed.
Kubernetes has the potential to fulfill unkept promises of cloud computing. While most IT organizations today employ multiple clouds, each of those platforms is managed in isolation. Instances running on multiple clouds should make it easier to create heterogeneous cloud computing environments that are centrally managed. In fact, Raghuram predicts that what was once viewed as several isolated instances of cloud computing before will be viewed simply as a common Kubernetes experience. Once that occurs, the whole need for the term “cloud computing” might dissipate as well.
In the meantime, as adoption of Kubernetes in production environments continues to expand, it’s becoming apparent that the balance for control over pricing might soon be shifting away from cloud service providers.