Google Compute Engine Is No Threat to Amazon, Microsoft, RackSpace

With limited configurations, virtual machine and operating system support, the beta version of Compute Engine isn't likely to steal users from Amazon and other IaaS leaders.

Mike Fratto

June 29, 2012

4 Min Read
Network Computing logo

Google has announced its IaaS service, Compute Engine, confirming rumors the search giant was getting into the cloud space. However, the limited number of VM, storage and networking options, coupled with the trial launch, means Google Compute Engine isn't yet a serious IaaS competitor.

Compute Engine is currently in preview phase; Google is accepting limited access for new accounts. Today, you can create a Linux VM consisting of one, two, four or eight cores each, with 3.75 Gbytes of RAM per core. The VMs are either Ubuntu 12.04 or CentOS 6.2, 64-bit Linux OSes. The networking options are thin as well, with what looks like Layer 3 networking, only with the ability to have a static or ephemeral public IP address and a firewall to set access control.

There are a number of options for storage. Each VM can have an ephemeral disk for temporary storage while it's running and the data written to it is encrypted. Once the VM stops, the ephemeral disk is destroyed. Persistent disks, also encrypted, can be attached to one or more VMs (though disks attached to more than one VM are read-only), and Google promises the same read and write performance of a local disk. Additionally, Google Compute Engine (GCE) VMs can access Google Cloud Storage.

"The GCE options seem to be a very random set of variables and one which might have flown five years ago," says Ben Kepes, an analyst at Diversity Analysis. "But my belief is that cloud customers have moved on and want far more flexibility and maturity in a platform."

The initial version of Google Compute Engine won't take share from other IaaS providers in the cloud computing space like Amazon, Microsoft or Rackspace, which offer a wider variety of configurations that can support OSes that require more capacity like Windows Server.

Google's own developer documentation states, "Compute Engine is designed to run high-performance, computationally intensive virtual compute clusters in the Google Cloud," which is clearly evident by the choice and configuration of the initial offerings for virtual machines and storage. The documentation also indicates GCE can be used in conjunction with Google's PaaS offering, App Engine. For example, App Engine could hold the user-facing components while Compute Engine performs computationally expensive operations like media rendering.

Next: Challenges Facing Google Compute EngineThe big question is whether enterprises will flock to GCE.

"Larger enterprises have proven willing, on the whole, to adopt cloud computing but will be increasingly cautious with a product that comes from a company that has a pedigree of making (and then ending) experimental products and whose very revenue stream comes from data mining customer data," says Kepes. While there is sure to be a strong separation between Google's cash-cow operations--search--and Compute Engine, enterprises will want to carefully review the license terms before using GCE.

Google will have to convince enterprises that it will both protect their privacy and remain committed to maintaining Compute Engine. While Google has gotten some big wins with other enterprise offerings like Gmail and Docs (the National Oceanic and Atmospheric Administration is using Google Apps and GM is reportedly moving to Gmail), there's still the question of getting adequate support when problems arise.

While support may be an issue, management should be less of one. Out of the gate Google has attracted a few key cloud management vendors, including Opscode Chef, Puppet Labs Puppet and RightScale. The integration allows centralized management and provisioning of the Compute Engine VM with the management frameworks and will help speed the deployment of Compute Engine resources. If Compute Engine picks up customers, other cloud management vendors like enStratus will likely add support.

The pricing during the trial period seems comparable with Amazon's EC2. An EC2 medium instance with 3.75 Gbytes of RAM, one virtual core and 410 Gbytes of storage is 16 cents per hour for a Linux image, compared with Google's entry VM, which has one virtual core, 3.75 Gbytes RAM and 420 Gbytes of disk space for 14.5 cents per hour. The data transfer rates are similar, as are storage and IP address charges.

The biggest differentiator between Google Compute Engine and others like Amazon, Microsoft and Rackspace is that the Compute Engine is in a trial phase, while the others are shipping services today. Google tends to take months or years to bring a service out for public use, so I don't expect its competitors to be worried yet. In fact, unless Google can find a way to differentiate itself from the others, Google Compute Engine is destined to be yet another cloud service.

About the Author(s)

Mike Fratto

Former Network Computing Editor

SUBSCRIBE TO OUR NEWSLETTER
Stay informed! Sign up to get expert advice and insight delivered direct to your inbox
More Insights