Skip to content

Robin Automation Platform for Creation of Kubernetes Clusters at Scale from Bare Metal Machines

Blog | 09.01.2020 | 3 min read

Sree Nandan Atur, System Designer

Sree Nandan Atur is a system designer and hands-on coder currently tackling the challenges of orchestrating multi-cluster and data center workloads at scale on Kubernetes at Robin.io. His vast experience spans various layers of software stack from firmware, drivers, system software, orchestration, chat bots, distributed systems and kubernetes. Prior to Robin, he was a member of the core architecture team at Tintri and built Tintri HyperV Management Software, Tintri Chat-bot and Tintri Analytics. Before that, he was at Microsoft Core OS Group building kernel drivers and system software for Windows 8. Nandan holds an MS in Computer Science from University of Florida, Gainesville.

  • Share on Twitter
  • Share on Linked In
  • Share on Facebook
  • Email this blog post

Recently, I had the opportunity to participate in a webinar with Srinivasa R Addepalli from Intel, and we discussed the creation of Kubernetes clusters from bare metal machines.  In this blog, I’d like to offer a few additional thoughts.

As we discussed in the session, 5G disaggregated infrastructure requires web scale management of highly distributed compute clusters. Due to the location and distribution of 5G radio units, massive amounts of compute clusters are deployed at geographically distributed nodes. These clusters need an orchestration layer that can discover, deploy and manage their life cycle operations at scale. Compute clusters formed out of bare metal machines often host large numbers of workloads, including VNFs, CNFs, CNAs and Network Functions. While Kubernetes has become the de facto platform for hosting applications which span across multiple clusters, there remains a need for multi-cluster infrastructure and service orchestration that can automate and manage day-0 and day-n lifecycle operations at scale.

To meet such need, the Robin Automation Platform combines simplicity, usability, performance and scale with a modern UI to provide bare metal, cluster, RAN and Application-as-a-service for both infrastructure and service orchestration. With Robin Bare Metal-as-a-Service, hundreds of thousands of bare metal servers can be provisioned with specific BIOS, firmware, OS and other software packages or configurations depending on the needs of the application – all with a click of a button. With Robin, it is equally easy to manage upgrades, as well as a wide array of hardware/server vendors, firmware, OS, software and Virtualized RAN (vRAN) and container platforms.

Using Robin K8S-as-a-Service (K8aaS), several tens of thousands of Kubernetes (K8S)  clusters can be managed along with their associated lifecycle management operations like instantiation, scale-out, scale-in, upgrades and termination, via a multi-tenant administrative graphical dashboard.

With Robin RAN-as-a-Service (RANaaS) functionality in the Robin Automation Platform, as many as tens of thousands of vRANs from multiple vendors can be orchestrated across multiple Kubernetes clusters. The initial Kubernetes cluster configuration requirements for network function packages, docker registries, and image caching policies are transparent to the user. In all cases, instantiation, termination, scaling, upgrades and healing are just an API call away.

The 5G stack needs other applications like large databases, load balancers, object storage, artifactory, vaults, caching/messaging platforms, etc. Robin’s cloud-native infrastructure of its Cloud Native Platform (CNP) runs these applications on Kubernetes. The Robin CNP is the ideal platform for hosting both stateful and stateless applications as it supports Virtual Machines (VM), LXC-based and Docker-based applications. It also includes a storage layer that provides data services, including snapshots, clones, backup/restore, and replication. The Robin CNP also includes a Telco-grade CNI that supports OVS, Multus, SRIOV and Calico.

 

 

Related Posts​