A stream of Google cloud executives took the stage in San Francisco on Tuesday to announce a slew of new cloud computing features, including around advanced network and resource-management capabilities. Speaking on a panel at the AppDynamics user conference in Las Vegas, Allan Naim, global product lead for the Google Cloud Platform, told the audience about how the company has been laying the groundwork for these new capabilities for years.
He explained how application containers can be a huge source of operational efficiency if you know how to manage them. Everything at [company]Google[/company] runs on containers and the company spins up 2 billion of them per week, he said. In the name of squeezing every ounce of capacity out of each server, a single box might contain hundreds of containers split among multiple workloads (Gmail and MapReduce jobs, for example) and it’s management software like Google’s Omega that helps ensure they’re all getting the resources they need.
In June, Google released its open source Kubernetes framework for managing Docker containers. It quickly gained adoption by IBM, Red Hat and Microsoft, which quickly released a set of Kubernetes-based services. It was only a matter of time until Google Container Engine came to be.
On the networking front, Naim said Google has spent a lot of time and money building out its software-defined networking infrastructure, which is now making its way into Google’s cloud service via the Andromeda framework. If that’s not enough to ensure network performance, he also noted that Google’s data centers are connected via fiberoptic network and the company has its own fiber running underneath both the Atlantic and Pacific Oceans.
“Not all clouds are created equal,” Naim said.
Of course, Microsoft and Amazon Web Services might take issue with Google’s claims of infrastructural dominance. One of the main talking points of Microsoft’s recent cloud roadshow featuring CEO Satya Nadella was the scale of its operation, which now spans 19 regions globally. Microsoft, like Google, Amazon and Facebook, also builds its own servers and puts a lot of thought into data center design.
And AWS, as General Manager Bill Platt pointed out in an earlier talk at the AppDynamics show, knows better than anyone the type of scale necessary on all fronts to handle millions of users — because it’s the cloud provider that currently has most of them. He repeated an oft-cited statistic that Amazon now deploys as much infrastructure every day as what it took to power Amazon.com when it was a $7 billion business. AWS ingests several million terabytes of data per minute and archives petabytes of data per day.
But so far, Google has been among the fastest cloud providers at turning its infrastructure and internal practices into products. For better or worse, the company has a very distinctive view of the way applications should be built, deployed and managed, and it shows in its cloud.