Guest: Madhura Maskasky (LinkedIn)
Company: Platform9 (Twitter)
Show: Let’s Talk

Platform9 helps enterprises run, scale, and optimize Kubernetes. In fact, it pioneered Kubernetes-as-a-service so companies can experience Kubernetes while leveraging the private cloud infrastructure, which is much cheaper.

In this episode of TFiR: Let’s Talk recorded at the KubeCon in Chicago, Platform9 Co-Founder and VP of Product Madhura Maskasky shares her insights on the cloud-ready Kubernetes landscape and how Platform9’s new EMP product will help companies with their cost optimization initiatives, particularly in the era of AI/ML.

How Platform9 helps enterprises:

• Platform9 wants to take away the Kubernetes deployment and management pain points, while not tying you to the public cloud resources only. It pioneered deploying Kubernetes-as-a-service so companies can experience Kubernetes while leveraging the private cloud infrastructure, which is much cheaper.
• If running Kubernetes on premises, on your own data center, colocation-hosted data center, Platform9 helps bring up that entire stack all the way from bare metal to virtualization layer to then Kubernetes. If you’re running AI/ML workloads, that can become one of the most cost-effective ways of running Kubernetes on premises without any of the Kubernetes operational burden.

• Its primary areas of focus are: 1) Solving operational complexity. Data scientists working on AI/ML especially do not want to deal with that complexity. 2) Cost becomes an even more important consideration because the amount of resources needed to run inferencing, the training algorithms, etc. is 10x or 100x more. If you’re running those resources in the public cloud GPUs, they are extremely expensive and in short supply.

What sets Platform9 apart from other companies offering cost optimization:

• It offers a fundamentally different approach. Its roots as a team are from virtualization. So, to reduce costs, instead of creating 2 instances as worker nodes for Kubernetes, Platform9 goes straight to AWS bare metal and deploys its own virtualization on top of it. It is able to spread capacity across a pool of bare metal servers. It can live-migrate VMs from one server to the other so that SLA is never compromised. Nobody else is doing this.

On its new product Elastic Machine Pool:

• Platform9’s Elastic Machine Pool (EMP) focuses on public clouds, specifically EKS. Kubernetes is not a very effective consumer of resources. Kubernetes clusters tend to be between 15 to 30%. Platform9 helps improve that utilization and shrink costs by 50%.
• It is a green solution. It provides a system-level and automation-based approach that lets you make better use of the capacity you have deployed, which is better for the environment.
• One of the key components of Platform9’s EMP is Rebalance. It’s a patent-pending component that uses data about your existing workload patterns to figure out the best ways to optimize cost or where to best place those workloads. That’s a place where generative AI could play a really big value.

What’s ahead for Kubernetes:

• In Kubernetes today, there is no native way to properly expose GPU resources in a way that enables sharing of those GPUs across spatial parameters and across time slicing. That’s an area of innovation that is coming to Kubernetes.
• There is talk about power optimization, which makes private and public data centers greener and also contributes towards costs.
• Scale is going to be taken to the next level with these newfangled AI/ML workloads.
• Kubernetes multi-cluster is still not solved and there isn’t a management cluster that can farm our workloads to a set of other slave clusters, per se. There are projects that will focus on this.
• There will be more innovation around simplifying the consumption of Kubernetes (i.e., using serverless components, not having to deal with VMs, simplifying constructs, etc.).

What’s next for Platform9:

• While its EMP product starts out with EKS, it is not limited to AWS or EKS. The technology has much wider applications across hyperscalers, which will be explored next.
• It will invest in AI/ML. There is a huge potential in enabling enterprises to make good use of GPU capacity across private and public clouds.
• There are MSPs that are getting a lot of customer traction from being able to host training workloads, etc., but they don’t have expertise to provide Kubernetes. Platform9 is looking to partner with them and provide its Kubernetes layer on top of their hardware. The end user benefits in terms of simplified complexity.

This summary was written by Camille Gregory.