Google is again taking aim at Amazon Web Services in the on-going cloud provider pricing war, revealing across-the-board...
cost reductions to its Google Compute Engine platform.
In addition to the price reductions -- which come roughly six months after Google last slashed its infrastructure as a service (IaaS) pricing -- Google unveiled a new virtual machine instance type it says will drive further cost savings for users.
In the new Google Compute Engine pricing model, Google's standard virtual machine types cost 20% less; its high-memory machine types cost 15% less; its high-CPU machine types cost 5% less; its small machine types cost 15% less; and its micro machine types cost 30% less.
Under the new structure, the typical cost for Google's standard instances range from one core and 3.75 GB of memory for $0.038 per hour to 32 cores and 120 GB of memory for $1.216 per hour. Before the cuts, these same instances ranged from $0.049 per hour to $1.538 per hour.
The price reductions are the latest in a series made by Google and rivals such as Amazon Web Services (AWS), as they battle for customer mindshare in the competitive public cloud market. In October, Google slashed its cloud instance pricing by 10%, prompting AWS to respond with various data transfer price cuts of its own.
It's likely this race to the bottom -- especially among the larger IaaS players like Google and Amazon -- will continue, said Robert Mahowald, vice president of SaaS and cloud services at analyst firm IDC in Framingham, Mass. Unlike some of the smaller or more niche IaaS vendors, who differentiate themselves based on value-added services, Google and AWS are more "hands-off" and base the bulk of their business on providing cloud infrastructure to ISVs. This, Mahowald said, makes pricing critical.
"[Google and Amazon] will have to compete, dime for dime," he said.
Because pricing and discount models vary between cloud providers, it's difficult to gauge the long-term costs of different IaaS platforms. However, Google's latest round of cuts does allow its instance pricing, at least in some cases, to dip below that of AWS. For example, Google's smallest IaaS instance, the f1-micro, is currently priced at $0.006 per hour, while AWS' smallest and cheapest machine, the t2.micro, costs $0.013.
Google's price cuts follow its Moore's Law pricing philosophy. According to the vendor, Google Compute Engine VM prices have been reduced by more than half since the platform launched in November 2013.
"It's our pricing philosophy that, every time we identify a place where we can make a substantial improvement in our own [infrastructure] costs, we should pass that on to our customers," said Miles Ward, head of solutions for Google Cloud Platform.
To attract new ISV customers to their platforms moving forward, AWS and Google should shift their focus from cutting prices to creating more long-term and fixed infrastructure pricing models, Mahowald said.
"I think they are waiting until they get more ISV customer types on board, but it's kind of a chicken and an egg thing – you're not going to get them until you can guarantee a fixed cost for their infrastructure for a year," Mahowald said. "All IaaS companies need to get their arms around being able to offer those kinds of models."
Google's new Preemtible VMs
In addition to the price cuts, Google introduced Preemtible VMs, a new instance type designed for short-duration batch jobs. Similar to Amazon EC2 Spot Instances, Preemtible VMs use Google compute resources that would otherwise be idle. And because these VMs are based on spare capacity and are limited to 24-hour runtimes, Google offers them for 70% less than it does a standard VM.
However, because Preemtible VMs "are designed to be interrupted," Google's sustained usage discounts cannot be applied to them, Ward said.
The new Preemtible VMs are designed for tasks that aren't time-sensitive, such as batch processing or rendering. Specific uses include video encoding, visual effects rendering and data analytics, according to Google.
Mark Johnson, CEO and co-founder of Descartes Labs, an image analysis startup based in Los Alamos, New Mexico, said his company uses Google's Preemtible VMs to process and analyze massive amounts of satellite imagery from NASA.
"There is nothing time-critical that we needed, so Preemtible VMs were a great way for us to reduce our costs by well over half," Johnson said. "For a small, seed-funded startup like us, that was really critical."
Johnson described the Preemtible VMs as being "like a contract with Google."
"You say, 'hey, I need this computing. When you have spare cycles, please let me use them.' It's good for batch processing jobs or very, very parallel jobs," he said.
The Preemtible VMs are in beta, with no word yet on general availability. The pricing model follows a fixed, flat rate that users pay by the hour – something Ward said sets Google's Preemtible VMs apart from competitive services that have "deeply variable" pricing models.
Kristin Knapp is site editor for SearchCloudComputing. Contact her at firstname.lastname@example.org or follow @kknapp86 on Twitter.