Intel, CoreOS Aim to Make Your Datacenter as Smart as Google’s

The goal of the joint project is to build a hyperscale-Cloud architecture with automated containers and appliances:

Intel has announced an effort with groundbreaking container developer CoreOS to design appliances that could make any corporate datacenter as smart as Google’s.

The two companies announced a joint project Intel officials described as an effort to democratize the sophisticated management and automation of hyperscale Cloud-computing environments by combining CoreOS containers with Google’s open-source infrastructure-management software Kubernetes, running preconfigured on appliances delivered by Intel OEM partners.

The result should be to make available to ordinary companies the efficiency, flexibility and performance demonstrated by the automated, virtualized “orchestrated container concept” that underpins hyperscale datacenters at Google, Facebook and elsewhere, according to an announcement of the partnership in a recent blog by Jonathan Donaldson, VP and general manager of Intel’s Software Defined Infrastructure group.

“At Intel, we are excited to be a part of the broad ecosystem that is working to bring containers to the masses,” Donaldson wrote in his announcement of the deal. “This is a key component of our desire to see private and hybrid Cloud computing grow significantly over next few years.”

CoreOS is packaging its container technology with a commercial version of Kubernetes – a systems-software management tool developed by Google to make it possible to run clusters of hundreds or thousands of containers from CoreOS rival Dockers as if they were a single system.

Kubernetes is designed to take over task scheduling and management for Docker containers, handle load balancing, provisioning and orchestration, and add self-healing functions to re-start failed processes, reschedule jobs, replicate containers to where they’re needed and group many containers into a single image of the application they represent to make it more easily addressable by users and other systems, according to Google’s Kubernetes documentation on GitHub.

Kubernetes supports CoreOS, but relies on Docker’s management infrastructure to download images into and run the containers that make up the clusters that make up the nodes that make up the services Kubernetes controls.

Support for API functions are spread around among management nodes to put interfaces to applications close to the hardware on which they run, but, so far, the master Controller Manager Server that controls the whole architecture is still centralized and run outside of containers, making the whole structure more hierarchical and dependent on top-down management to be as efficient as it could be, according to the Kubernetes site.

CoreOS bundled its Linux-based application container – which is built on the open-source version of Docker – with Kubernetes to get Tectonic — which packages all the components for a  Google-like Kubernetes-controlled Cloud infrastructure, but adds dashboards, workflow tools, registries to connect shared Linux containers and other refinements designed to fit more neatly into existing corporate datacenters.

In that effort CoreOS, which refers to Tectonic as “Google’s infrastructure for everyone else,” launched Tectonic April 6 with the help of a $12 million investment from Google Ventures, which expects to see enterprise computing mirror the techniques and structures of hyperscale cloud providers, according to a  statement in the announcement from Google Ventures partner Dave Munichiello.

Google’s goal for Kubernetes is to have it run on all Cloud platforms, not just Google’s, according to Google’s Kubernetes product lead Craig McLuckie in the same announcement.

“Our Google Cloud Platform customers benefit from modern management patterns pioneered in Google, and Tectonic frees them to pick their Cloud provider solely on the merits of their infrastructure, with no lock-in whatsoever,” he said.

Intel’s goal is to help CoreOS scale tectonics quickly and package it conveniently enough to be rolled out easily and efficiently to traditional datacenters, Donaldson wrote in his blog.

One good way to do that is to help OEMs build the whole package into highly modular datacenter server hardware customers can use to easily build datacenter-based Cloud servers that work like Google’s but are far easier to manage.

“We expect that this work will lead to ready-to-ship, hyperscale Cloud systems that coincide with the future GA release of the Tectonic suite,” Donaldson wrote.

Intel has not delivered a full set of reference specifications for the Tectonics bundle running on Intel hardware, or announced pricing for servers running it. It has announced that two companies – modular-server maker Supermicroand integrator Redapt – would sell Intel-powered versions of Tectonic servers.

The Tectonic prototypes have already gotten a 2X boost in performance using asynchronous DRAM self-refresh – a feature Intel added to Xeon E5-2400 v2 chips in 2014 – according to a May 5 post from CoreOS CEOAlex Polvi that described the effort.

“This is the beginning of a deep partnership to enable businesses to take advantage of containers, distributed systems, and the next generation of infrastructure,” Polvi wrote.