Our India data centre should come up this year: Google's Mohit Pande

Q. Give us an overview of Google Cloud for enterprise customers.
Google Cloud is the umbrella brand for all our cloud services. From an enterprise perspective, there are four pillars. The first is G Suite comprising Gmail, Google Drive and collaboration software for businesses. Globally, we have over 3 million businesses paying to use G Suite.

The second is the Google Cloud Platform—comprising infrastructure [servers, storage, networks], platforms, analytics and machine learning. The third is Google Maps, which consumers are familiar with. Also, a lot of businesses use it to run location-based operations—asset tracking, route optimisation and even business intelligence. For example, financial services have used Maps to visualise concentrations of high-net-worth customers.

The fourth is Devices, which includes Chromebook laptops, Chromebox for videoconferencing at meetings, Chrome for signages, and Android for enterprises to manage and secure devices.

We’ve been investing significantly in the Google Cloud Platform—close to $30 billion—as capex to build infrastructure, in the form of cloud regions, or data centres, over the last 2-3 years. We are investing a lot in the Asia-Pacific region and in India. We have announced an India region, which should be coming up by the end of this year in Mumbai. We are also investing in what we call Google’s Premium Network, which includes undersea cables and so on to reduce latency, and increase reliability and security.

Q. In addition to Android, Google has also put many other projects in the open community, which are relevant to cloud computing. Could you give us a few examples?
One of the two strong examples is Kubernetes. It is a framework for deploying what are called application containers on the cloud—describe the processing power, storage and networking needed to run a software application, and a [virtual] container will pull together those resources from the cloud. Today, Kubernetes has become almost like a standard for deploying and managing application containers on the cloud.

The second example, on the machine learning side, is TensorFlow, which is a versatile machine learning library and AI engine that Google open-sourced two years ago. It allows even people with little or no knowledge of machine learning to plug in their own datasets and solve problems. A case study on Google’s website shows a Japanese farmer using TensorFlow to sort cucumbers on the basis of shape, size, colour, number of prickles and so on.

Q. Cloud computing has evolved from just an infrastructure play to more complex capabilities being provided as a service. Many of these have been built by Google...
A standout feature of Google’s cloud is the focus on analytics and machine learning. We have been working with large amounts of data which has led to a lot of technologies built within Google, first to manage our own business and then offered as services. The number of options around data storage and management that we offer is quite large.

Take Cloud Spanner, which is a massively scalable database. It can handle millions of transactions a minute. Consider BigQuery, a data warehouse on the cloud. With BigQuery, all you need to do is get your data. You don’t have to worry about the complex underlying infrastructure and, at incredible response times and return on investment, you immediately start getting useful information out of the data.



Q. How are advances in machine learning yielding plug-and-play cloud products that businesses can use today?
Machine learning is another area of differentiation and Google started at home, building its own products with an AI-first approach, from spam control to prioritised mail and smart replies and recommendation engines in YouTube and the Google app. Google is now offering a lot of these capabilities to its customers so that they can build their own machine learning-based applications. And we offer the entire spectrum—from pre-trained APIs [application programming interfaces that allow different applications to talk to each other], which means customers don’t have to build their own models, to Cloud ML, where customers can build their own models.

Translate, speech-to-text, natural language processing, vision, video intelligence…these are areas where Google has used large public data sets and its expertise in AI algorithms and its own hardware to build plug-and-play products in machine learning for businesses.

On hardware, Google has announced the second generation of its TPUs, or tensor processing units, which are specialised hardware for machine learning algorithms for training AI programs and other applications. These are already available on Google Cloud.

Cloud ML is being used in multiple industry verticals. Financial services companies are using it for fraud detection, retail chains are applying it to determine the next logical purchases of consumers and oil exploration and weather forecasters are using it for their modelling and so on.
 
On the business front, we have been innovating on how customers are charged, with options such as per-minute billing, and custom virtual machines, which allows customers to configure the number of processors they want to use and so on. Sustained usage discount on Google Cloud automatically shifts customers to more attractive rates as usage increases.

Q. How will opening a local data centre in India help?
India is a very strategic market for Google; there’s a lot of potential across different customer segments. Opening a region in India allows Google to serve Indian customers from closer quarters and improve a host of service parameters. Latency, a measure of the time it takes for data to travel back and forth, can be reduced, for example.

Another worthwhile consequence will also be that a data centre locally will spur the growth of partnerships with developers, independent software vendors, system integrators and so on. As India improves overall with its internet connectivity, smartphone use and so on, businesses will see the increasing need to tap cloud computing.