There is a misperception about the new Google Cloud Platform that the company put into general availability last week at Google I/O. It’s not a brand new platform. It’s what Google has used for years. It is Google’s foundation. It is what makes Google, Google. And now it’s open for the first time to developers and businesses.
Google Platform is new in the sense that anyone can now use it. But until now only a relative few number of people have had access to the platform.
Google Cloud Platform officially launched at last year’s Google I/O. So it still has a lot of hype that comes with a new Google service, especially at an event like Google I/O. It does not have the full set of features that comes with Amazon Web Services (AWS). A customer can get a much deeper service level agreement (SLA) from Windows Azure. Customers can use a platform-as-a-service (PaaS) like Openshift and leverage the Red Hat infrastructure. OpenStack is an option for companies that want to build out their own open cloud environment. Go that route and a customer has a host of vendors to choose from. Red Hat, IBM and HP are just a few to choose from for any number of software and services.
The Power Is In The Network
But there is one thing in particular that sets the Google Cloud Platform apart. And that’s the network that connects the company’s data centers so questions can be answered in milliseconds. It’s what makes it possible for Google to offer 3D maps, translation APIs and Google Glass.
“It is blazing fast,” said Will Shulman, co-founder of MongoLab about the network in a panel at Google I/O about distributed databases. “The other thing – it has a private distributed backbone between all the data centers.You are talking over Google’s backbone, not over the Internet.”
The network speed makes a difference in a few ways. The compute and storage in Google Compute Engine are separated but for the user it appears as if it is all together because it is so fast. It’s like having one giant, programmable super computer that in reality is distributed across thousands of servers.
The network speed also helps make a difference in cost. With the speed, comes the ability to process more data in less time.
Google factors its network into its pricing, much like cloud provider ProfitBricks does. ProfitBricks uses InfniBand, which offers more bandwidth capably than Google’s 10 gigabyte network. Regardless, Google’s fiber network and data center optimization provides the opportunity to offer sub-hour pricing, down to the minute.
On the Google platform, a customer can double the cores and do a data job in 30 minutes at the cost that it would normally take an hour to do.
Google views data centers as living things. They are not islands but exist in a connected world, connected to devices, other services and other data centers.
It’s this view that shows why Google has to be so considerate of its own network. The world is becoming a vast data fabric. But networking is expensive. Compute and storage costs continue to decrease but networking has not gone down at the same pace as CPU and storage, said Google Product Maanger Amit Argawal in a presentation at the Open Network Summit last June.
What it costs to connect a 10 gigabyte pipe between two regions in the United States is different from connecting different countries in Asia, where the markets are emerging fastest, In the video, Argawal says in the video. Devices are ubiquitous and disposable. Someone can lose a smartphone, buy a new one and be back up in a half-hour. The data is in the cloud not on the device. The services in turn are populating across the network. Put together it’s a virtuous circle. The network needs to be fast and interactive. If not, user engagement will slow. High availability needs to be built into all layers of the stack.
Why Developers Play A Crucial Role
To allay networking and other costs, Google has to continually keep its operations running optimally. The Internet business model means services have to be free or for a small fee. That means Google has to make sure developers are building apps on services that will help Google extend its advertising products and low-cost cost subscription services such as Google Apps.
And that’s why Google Cloud Platform plays an important role in attracting more developers, who in turn help extend Google’s properties.
For example, Google talked at Google I/O about how it offers tools to help developers integrate into the Google back-end. Google Maps, Chrome. Android and BigQuery all have these integrations. Google Glass will get integrated but for now it is not the number one focus.
AWS has a rich developer ecosystem and has a deep selection of services to offer. But Amazon is not an identity and services provider like Google is. Google has more data to offer developers so that will also be a strong selling point going forward for the company with developers.
For Cloudant, a distributed database company, it’s the fact that there is now another community outside AWS that it can tap. “There are a large and growing number of developers on Google,” said Co-Founder and Chief Scientist Mike Miller, who also sat on the distributed database panel.
Google App Engine symbolizes some of the differences that may attract developers. Google announced at Google I/O that PHP would be offered on Google AppEngine. This will make Google available to the scores of web developers who have built their web sites with the programming language. In March Google acquired Taleria, showing its continued emphasis on building out support for dynamic programming languages and need for systems that scale out efficiently. From Frederic Lardinois post about the acquisition:
The company claimed that its technology allowed developers to “handle more users with fewer boxes, without changing a line of code.” Talaria also claimed its ” server lets you keep your favorite high-productivity languages, but with the scalability and performance you’d expect from a compiled language.”
And then there is the ease of use that Google is trying to offer with Google App Engine. These include back-end as a service tools and more management features that allow developers to focus more on the code then the back-end.
That’s important for companies such as OrangeScape, a “visual PaaS,” for non-developers to build apps. CEO Suresh Sambandam said that means the company can keep its IT team relatively tight.
Google has a network that makes it arguably one of the largest carriers in the world. But it’s the cost of these data centers that will be its biggest challenge going forward. It’s almost as if Google had to open its infrastructure to extend its distributed network as efficiently as possible while continually attracting developers to scale its business model.