Google

Google's master cloud plan: Buy more infrastructure, charge less for it

Google's Miles Ward tells us how Google plans to let its cloud customers run like Google... for less.

Google cloud

Google is not yet top-of-mind for enterprises looking to spend big on the public cloud, with Google claiming just 5% of the cloud infrastructure market, compared to Amazon Web Services' (AWS) 28% and Microsoft's 10%. Rather than settle for also-ran status, Google has a strategy for up-ending the cloud market to its advantage.

The game plan? Spend big, charge less.

How much more is Google spending? In 2014, Google spent double what Microsoft and Amazon spent on cloud infrastructure ($10.9 billion vs. $5.3 billion and $4.9 billion, respectively). And yet, according to Google's Miles Ward, Google also charges less. Between those investments and the pricing wars, Google's cloud business has jumped 101% year-over-year, according to Technology Business Research analyst Jillian Mirandi.

That's not all infrastructure spend, but it still shows the cloud challenger is making progress. That progress, as Ward (who spent years building the AWS cloud business) told me in an interview, will continue as Google's "core technical investments... will pay dividends to our customers over time," allowing them to "run like Google."

TechRepublic: You've publicly claimed that Google's cloud comes in lower than AWS. Is that always the case?

Ward: Yes, always. As long as you're doing the math right — accounting for cost of capital, equivalent instances, etc. — no matter which selection of Reserved Instance (RI) on AWS you pick, Google Compute Engine has a price advantage over EC2 today. Many other Google Cloud Platform services enjoy the same cost saving, and many customers I've worked with are saving money on their complete infrastructure by moving to Google Cloud Platform.

For example, Starmaker recently talked about saving $3,000 a month based on per-minute billing alone. Our Sustained Use Discounts, which kick in automatically without any commitment or up-front payment, extend that advantage further.

I wrote a long blog post that breaks out a step-by-step analysis of an example workload, demonstrating that for steady or burst, big or small, paid up-front or not, we're the price leader; and we don't make you jump through hoops or spend precious capital.

TechRepublic: Assuming the math works out, does it matter? That is, even if you're a few percentage points cheaper, is cost the primary driver of cloud?

Ward: A cost advantage of single digit percentages likely won't drive average users to switch providers, as switching costs today are still too high. But we hope it will encourage developers to start out on the price leader, which, as I've said, is Google Cloud Platform.

This narrow difference also assumes that if you're using AWS, you are using RI's, which is far from a majority of users. If an AWS customer is running ~24/7 and is not on an RI, the price gap is very substantial and very likely worth the effort to switch. It worked for Atomic Fiction, which started on AWS but has since jumped on the Google Cloud Platform.

TechRepublic: How much do companies think about cost when choosing a provider, and how should they?

Ward: Customers ask me about cost all the time, so I think it's definitely top of mind.

Predictability is also critical to businesses, as are quotas and controls. Google App Engine has had quotas for years!

I helped build the pricing calculator for Google Cloud Platform, but you still need real information about your workload to be able to estimate the cost. In most cases, it's simply most accurate to do a proof of concept, run it, and extrapolate. Often folks can improve efficiency after they start, and load testing can help you estimate what more customers/users will do to your costs, but nothing beats a little dry run.

One of the things we're working on are some public "bill-studies," where we'll get permission from a customer to publish their full bill and analyze their usage pattern. This kind of detail will really help folks understand what to expect.

TechRepublic: As important as low pricing may be, flexibility strikes me as an even bigger selling point. In your experience with both AWS and Google, how does this on-paper flexibility advantage translate into real-world benefit for customers?

Ward: It's absolutely massive. The risks of low flexibility are precisely why most folks investigate cloud in the first place.

Why would you come to cloud only to lock yourself back into the same old inflexible long-term planning requirements? If we're trying to automate our infrastructure to make it flex to our needs, why would we accept inflexible pricing? We believe the goal is to make pricing as flexible as possible.

One counter-intuitive part of this is how totally predictable some loads or load characteristics really are: "guess what, when it's nighttime, fewer people use your app" or "Check it out, at the end of the month, you're going to run a bunch of batch analysis."

Despite these predictable patterns, you'll note the payment models don't tend to smoothly enable cheaper usage in these terms; you still have to automate, you still have to manage.

TechRepublic: OK, I get the cost and flexibility arguments, but AWS has a massive lead in functionality. Why should a company choose Google over AWS?

Ward: I can talk about this for days, but basically, there are core technical investments that Google has made, which will pay dividends to our customers over time. These investments include our network, our connectivity, our security technology, our investments in open source, our infrastructure tech, our development tools, and the enormous human capital in our expertise around distributed systems, machine learning, and familiarity with operations at scale.

When I talk with customers who are choosing Google, they're making a long-term choice rooted in the confidence that Google already has the best cloud, it just has to hurry up and let everyone else create amazing things with it.

Take a look at Kubernetes and Container Engine, Cloud Dataflow, or Andromeda: this stuff is giving every developer direct access to the tools Google uses to build Gmail, Apps, Search, Maps, and more.

Today's differentiators, like per-minute billing, outrageously fast app, instance, and load balancer scaling, more pay-per-use services vs. pay-for-time services? These are just leading indicators of the technical strength that will continue to enable us to release features that other clouds will have real difficulty matching.

I definitely think that "running like Google" is a goal many customers have. We're working double time to let them do even better than we do.

Also see

About Matt Asay

Matt Asay is a veteran technology columnist who has written for CNET, ReadWrite, and other tech media. Asay has also held a variety of executive roles with leading mobile and big data software companies.

Editor's Picks

Free Newsletters, In your Inbox