Wrapping Kubernetes Around Applications Old And New
February 16, 2017 Jeffrey Burt
Kubernetes, the software container management system born out of Google, has seen its popularity in the datacenter soar in recent years as datacenter admins look to gain greater control of highly distributed computing environments and to take advantage of the advantages that virtualization, containers, and other technologies offer.
Open sourced by Google three years ago, Kubernetes is derived from the Borg and Omega controllers that the search engine giant created for its own clusters and has become an important part of the management tool ecosystem that includes OpenStack, Mesos, and Docker Swarm. These all try to bring order to what at times can be a chaotic environment for hyperscalers and HPC centers and their analogues enterprises, as we have talked about at The Next Platform.
Two years ago, Univa CEO Gary Tyreman took a look at the options out there and decided the company’s future would be tied to Kubernetes. Univa’s core business has focused on Grid Engine, the open workload scheduler software designed to manage and orchestrate distributed HPC and enterprise computing environments, ensuring that workloads got the resources they need and in the order they need them. Univa took over support of Grid Engine in 2011, when Oracle let go of the program after inheriting it through its acquisition of Sun Microsystems a year earlier. Univa was able to bring in much of the Grid Engine engineering team from Oracle, says Tyreman.
More than two years ago, Grid Engine users began agitating for Univa to bring Docker container support to Grid Engine, with some so anxious that they began figuring out ways to make that happen on their own. Given that, Univa came out with Grid Engine Container Edition, which automates the process of deploying Docker Engine runtime for containers on top of clusters of servers. It was around the same time that Tyreman and his team made the decision that the company would jump on the Kubernetes bandwagon. It was the direction customers were taking, it had a lot of support in the open-source community, and was by far what people were talking most about.
“The momentum is there,” Tyreman says. “There is no question that it was the right choice. Everyone we talk to is talking about Kubernetes. They’re not talking about the second or third choice.”
In recent months, the focus for Univa has been making orchestration and management of mixed environments – where some workloads are being run in containers and others aren’t – easier and more unified. In similar fashion to datacenter infrastructures, applications are disparate group. While X86 servers dominate most datacenters, there are still IBM mainframes that continue to hum along running legacy programs that are mission critical and that organizations are unlikely to swap out anytime soon. The software environment is the same way, Tyreman said. In customer facilities, Univa not only sees Linux, but also IBM’s AIX, Hewlett Packard Enterprise’s HP-UX, and Solaris from Oracle. And many of those aren’t going away any time soon.
“Sure, there are a lot environments in enterprises where they’re not going to build a lot of new AIX apps, but they will still have an AIX environment,” he said. “It will be a truly mixed environment.”
It was against this backdrop that Univa in December released Navops Command, a workload scheduling and policy management tool that can run on any Kubernetes distribution, such as OpenShift from Red Hat and Tectonic from CoreOS. However, a key to Navops Command was the ability to run both shiny new containerized applications and older legacy (and uncontainerized) workloads within the same Kubernetes cluster, sharing the same compute, networking, and storage systems. The capability makes it easier for organizations to embrace containers and Kubernetes in their datacenters by removing the requirement for two environments – one for their microservices-based applications and another for those workloads that don’t run in containers. Mixed application environments aren’t going anywhere, and neither will the demand to run workloads in containers and to manage them with Kubernetes.
“We found some customers want to run in a mixed workload environment,” Cameron Brunner, chief architect of the Navops product suite, said during a video presentation outlining Navops Command. “Really what that means is they have some containerized apps that run inside of Kubernetes just fine, but they also have a lot of workloads that don’t necessarily run in Kubernetes, and they want to run those just as they have without really modifying anything.”
According to Brunner, Univa containerized Grid Engine, which can run under Kubernetes and essentially creates an HPC cluster hosted by Kubernetes. The Command software places the Grid Engine contents into a Grid Engine “pod,” which will then schedule HPC workloads that are inside of them. Through this, users are able to run both containerized and noncontainerized workloads within Kubernetes.
Tyreman said the ability to more easily orchestrate and manage mixed application environments will be a growing focus for organizations as they embrace containers and Kubernetes. The demand for this is beginning to gain steam, and Univa already has several proof-of-concepts up and running with customers. The company also is close to releasing the next version of the Command software.