The future of enterprise computing looks a lot like a cloud, and that cloud will increasingly run datacenter-level operating systems. Apache Mesos, born at UC Berkeley and embraced by Twitter, eBay and Airbnb, is coming soon to an enterprise near you.
Christos Kozyrakis, an associate professor at Stanford, is a rock star in academia research around datacenter scaling, security and quality of service. We’re talking Google scale. His research drew a lot of attention earlier this year, including a feature in the New York Times. So a lot of heads turned on Wednesday when he jumped from the Farm to the San Francisco-based startup Mesosphere.
I wrote about Mesosphere earlier this summer after it closed a $10 million Series A round led by Andreessen Horowitz. Mesosphere is a major contributor to Apache Mesos. The conceit is that Mesos is like the open source kernel of an enterprise-class scaleout platform that Mesosphere is building; one that functions much like an operating system for the modern datacenter.
I caught up with Christos to ask about his plans at Mesosphere.
ReadWrite: You’ve worked closely with a number of cool cloud startups as well as giants like Google. Surely you’ve been wooed by some of them. Why did you finally go to Mesosphere?
Christos Kozyrakis: Large-scale datacenters running private or public cloud services are the future of enterprise computing. This creates the need for an operating system that operates at the level of the datacenter, so that developers and operations teams can largely forget about individual servers.
Mesosphere is building this datacenter operating system. It’s a great opportunity for any systems researcher..
RW: How does your research fit into the Apache Mesos project and where Mesosphere is going?
CK: My research over the past seven years has focused more and more on management and scheduling algorithms that make datacenters faster, greener, and cheaper. Integrating these algorithms in Mesos is a great way to get them widely deployed and find the next set of opportunities for further advances.
RW: Your Quasar scheduling algorithm seems to have many parallels to what Mesosphere is doing with Mesos and what Google does with Borg/Omega/Kubernetes.
CK: There is great synergy. Mesos and Omega are datacenter operating systems that can benefit from a scheduler like Quasar that brings a big data approach to large-scale resource management. Kubernetes defines APIs that makes it easy to write service-based applications. I am excited to see the benefits from coupling these technologies.
RW: It’s the hot topic on cloud panels now, containers vs VMs. Some argue that VMs solved a problem a decade ago in datacenters, especially around server consolidation, but containers can do even more for modern datacenters being built today.
CK: Containers and VMs each have advantages for application deployment mechanisms, and one of the great things about Mesos is that it supports both. The issues I am most excited about – scaling applications to thousands of servers, raising utilization in datacenters, security and reliability at scale – can be brought to container- and VM-based environments using Mesos. Customers can make choices based on their specific use cases.
Lead image courtesy of Shutterstock