5 Comments

Summary:

Google has announced plans to start offering a compute-on-demand service that rivals Amazon’s EC2 service. The company will initially offer this as a limited preview to larger customers. It is a strong and logical move by the company as it continues to view for developer affection.

googlecloudplatform

Google has announced plans to start offering a compute-on-demand service that rivals Amazon’s Elastic Compute (EC2) service. Google has offered many what it calls higher-level cloud services such Google storage, BigQuery and Google App Engine in the past, but now the company believes it needs to sell a more prosaic Infrastructure as a Service (IAAS) offering whose primary target is attracting more developers to Google’s cloud platform. The news of this new service was first reported by my colleague Derrick Harris in May and was confirmed later by me with additional details this past week.

“The Google Compute Engine, we believe, has been the missing piece,” said, Urs Hölzle, Google’s senior vice president of technical infrastructure, during a broad conversation this week. He said that building an infrastructure-as-a-service isn’t a trivial task, as the demands on such a service are quite intensive. Google has been working on this new service for some time now, he added.

The focus of the Google Compute engine is on performance, scale and value. In order to show its performance and scale, Google is planning to show off a genomic app that runs on 600,000 cores. Another app will use 10,000 virtual machines. And if that isn’t enough, the company says it will offer 50 percent more compute resources compared to other shared cloud infrastructures. Translation: It’s a shot across the bow of Amazon Web Services’ EC2 offering. (See here for pricing)

The developers can run any stack and any software on this service. The company is partnering with third-party services such as RightScale to add more tools and services to its platform. Google is going to initially offer its service in limited preview and will sell it through its sales force if customers are looking for 100 or more cores. Eventually the service will be accessible with a credit card and a browser like most cloud services.  From Google’s blog post:

The capabilities of Google Compute Engine include:

  • Compute. Launch Linux VMs on-demand. 1, 2, 4 and 8 virtual core VMs are available with 3.75GB RAM per virtual core.
  • Storage. Store data on local disk, on our new persistent block device, or on our Internet-scale object store, Google Cloud Storage.
  • Network. Connect your VMs together using our high-performance network technology to form powerful compute clusters and manage connectivity to the internet with configurable firewalls.
  • Tooling. Configure and control your VMs via a scriptable command line tool or web UI. Or you can create your own dynamic management system using our API.

At launch, we have worked with a number of partners — such as RightScale, Puppet Labs, OpsCodeNumerateCliqr and MapR — to integrate their products with Google Compute Engine.  These partners offer management services that make it easy for you to move your applications to the cloud and between different cloud environments.

A company spokeswoman said that anyone can “sign up today, but we will be accepting customers who are focusing on larger workloads. In some cases we would accept smaller workloads as well. ” During the early phase, Google will offer Google compute service only to the U.S.-based developers, but will eventually roll out the platform to customers globally. Hölzle said that the company was using Google’s current infrastructure stack to offer the on-demand compute service.

Better late than never?

When asked if he believed that Google was a tad late to the party, Hölzle pointed out that while there have been many existing players offering cloud infrastructure services, there is ample opportunity for Google as the shift to cloud is more cyclical and long term. “This really isn’t about stealing marketing share from other players,” he said.

“I think we are early because the whole industry itself is in [its] infancy,” he said. “If you look at it, in the grand scheme of things, nearly 99 percent of the companies are not in the cloud.” Hölzle, however, said it was the right time for Google to enter the market. “More and more apps are being built for the web and mobile and the original storage and services are all moving to the cloud,” he said. The emergence of Chrome OS, Android and iPhone have led to the point where cloud clients are becoming “stateless.”

“It is very early in the market and, frankly, five years from now you will have a whole different kind of cloud and services.” He declined to outline what the cloud will look like in five years.

Hölzle was reticent about predicting the level of adoption as well, but was not shy of pointing out that Google has been in the infrastructure business for years and it is one of the key advantages for the company. “The market will show,” he said, and invited me to ask him the same question “two years from now.”  

The great (cloud) game

Despite Google’s dismissal, one can’t help but notice Amazon’s looming shadow on Google. Amazon Web Services, thanks to being an aggressive and early believer in the cloud as we know it, has carved itself a nice niche and is rumored to be bringing in over a billion dollars in revenue. But it is not just revenue that has a whole industry jealous of Amazon’s cloud business.

Success in cloud services has made Amazon attractive to startups and independent app developers, who are embracing Amazon’s stack of cloud services. These code-tinkerers are the kingmakers in this new world, especially now that Amazon has forked Android and has been pretty public about its grand mobile ambitions.

The battle for developers and locking them into cloud and mobile platforms is literally the trillion-dollar question of the 21st century. Microsoft Azure, Apple’s iCloud, Amazon Web Services and now Google Compute Engine are essentially trying to get their hooks into the developers. Frankly, I am surprised that Facebook hasn’t announced its own efforts to do the same.

Amazon, for now, is the king of the hill. At our Structure 2012 conference, when I asked Amazon CTO Werner Vogels about the next five years, he talked about a new layer of services emerging and Amazon being the trendsetter. It is a distinct advantage for the Seattle-based company that has angered its partners but has been focused on making sure it keeps pushing the envelope. He understands — and so does Google — that there is an opportunity to take away the dollars spent on IT dinosaurs such as Hewlett-Packard.

These giants of the past should be waking up with a migraine, for the entry of Google makes life tougher for them. I wonder what this news does to smaller cloud players such as Rackspace that have been inching their way toward Amazon’s heels.

Nevertheless, Amazon knows it has no time to rest on its laurels. For instance, it is not going to let Google press the price advantage for long. “If you look back, we’ve lowered pricing 20 times, so the best thing to look forward to is we’ll continue to do that,” Vogels said in our onstage conversation. “That’s at least our goal.”

And whichever way you look at it, Google’s entry into the business is a good thing for the developers and startups.

  1. Joyent is ¢4 per hour for the larger instances compared with ¢5.3 per hour at Google. And Joyent doesn’t have to nickel and dime customers with bandwidth charges. How can Joyent compete and win on price? Joyent has different technology that is better, faster and has a lower cost basis. By going with Linux based KVM with non-local “ephemeral” storage, Google has selected the wrong virtualization model. Telecoms like Telefonica are using Joyent’s technology to out-compete Google. The battle is going to get interesting. Just as Google Plus is still struggling against Facebook, Google Cloud Engine is not a guaranteed win in the IaaS space. That has broader implications for Google’s ability to capture the new generation of App developers build solutions stretch across phones, tables and the web. If those App developers are not using a Google Cloud, then they may not use Google payment engines, or Google advertising engines, even when part of their apps live on Android and iOS phones.

    Share
    1. Mohnish Chaudhary Friday, June 29, 2012

      Joyent’s least expensive plan (source Joyent website):

      SIZE DISK PRICE/HR
      Small 1GB (1 CPU) 30GB $0.085

      Share
      1. Rod Boothby Friday, July 6, 2012

        Joyent’s plans are all in. Joyent doesn’t charge in piece meal for every put and get, every bit of traffic. Bandwidth is free… up to 10TB per month, which is a huge amount.

        Share
    2. Rod Boothby = Vice President, Corporate Business Development at Joyent

      Share
  2. Its great to welcome Google to this horse race. Urs Hölzle is right about this being cyclical paradigm shift, not a fleeting fly-by-night shift. I’d love to see a live debate between Urs Hölzle, Microsoft’s Rolf Harms and Amazon’s Werner Vogels! The greatest news about all of this is that there is no monopoly! Ding dong the Wintel monopoly witch of the past is dead! Rejoice! Hurrah! My only beef with Google’s announcement today is that they are not making the debut of Compute Engine in such a way that is apt for startups, which is ironic given Google I/O is a developer conference. But in time its reasonable to expect decent developer support for companies of all sizes. The key to any of these big three vendors is to avoid “lock in”. Most people in the enterprise space today are not dumb or naive (unless they were born yesterday) so their will be very little tolerance for “Hotel California” like lock ins.

    Share

Comments have been disabled for this post