Google is offering for the first time the option of rolling your own server on the Google Compute Engine cloud.
Customers may configure a server with exactly the virtual CPU and memory that they want, as opposed to selecting one from a table of preconfigured instances, such as those offered by Amazon Web Services and other cloud providers. Google calls the option Custom Machine Type instances. It was made generally available Feb. 17.
Amazon has a wide selection of 39 different instance types, but it doesn't allow the customer to simply configure one the way he or she wants to.
Google Custom Machine Types were first introduced Nov. 18 and have been available through the Google Developer's Console as beta instances for the past four months. In announcing their general availability, a Google spokesman claimed not only are Custom Machines more flexible, but cheaper than their Amazon counterparts.
"When choosing a virtual machine type, major cloud providers force you to overbuy, since MVs usually come in powers of two. You need to buy 8 vCPUs, even when you only need 6. Today this ends. With Custom Machine Types, you can create virtual machines with the shapes (i.e. vCPU and memory) that are right for your workloads," wrote Google's Sami Igram, product manager for the Google Cloud Platform, in a November blog post.
With Custom Machine Types, a customer can create a server with from 1-32, virtual CPUs (the vCPUs must be assigned in even numbers) and assign it the amount of memory desired. Each virtual CPU can be given up to 6.5GBs of memory.
Google also offers pre-configured virtual servers, like other providers and offered an example of when a Custom server would be a better deal than what it calls its Standard server type. The n1-standard with eight vCPUs and 30GB of memory would cost $204.40 if it ran continuously over the course of a 730-hour month. The next size up, an n1-standard with 16 vCPUs and 60GB of memory, would cost $408.80 over the same period.
A Custom Machine Type sitting in the middle of these two with 12 vCPUs and 45GB of memory would cost $321.75. The hourly rates amount to $0.28, $0.441, and $0.56, respectively. By gearing the virtual machine to what a customer actually needs, 21% is shaved off his hourly rate with the Custom instance over the larger n1-standard instance.
By adopting the Custom approach, "our customers have saved an average of 19%," wrote Igram in a Feb. 17 blog post announcing general availability.
Google, like Amazon, charges for the amount of time a virtual machine runs, but it measures that time by the minute instead of the hour.
In a Jan. 8 blog post, Miles Ward, Google Cloud Platform global head of solutions, compared the price of a Custom Machine Type with two vCPUs and 8GB of RAM to the Amazon Web Services "M4 Standard," although there's no such instance in the AWS matrix chart. The basic M4 appears to be the M4 Large, with two virtual CPUs and 8GB of memory. The Custom costs $54.82 for a month's worth of use to the AWS M4 Large's $87.60.
But such comparisons are hard to make without a clear definition of what constitutes a virtual CPU for each, down to the gigahertz speed of the chip behind the virtual CPU.
[Want to learn about early 2016 AWS price cuts? Read Amazon Starts 2016 with three price cuts.]
"Rest assured Google continues to be the price/performance leader in public cloud," Ward wrote in January after an announcement by AWS of three price cuts. "As you can see, we're anywhere from 15-41% less expensive than AWS for compute resources ..."
Regardless of the exact price comparison, many customers will feel they are getting the better end of the deal if they configure what they want and pay for only what they configure. Google has instated an innovation in the cloud that Amazon may one day have to emulate.
Does your company offer the most rewarding place to work in IT? Do you know of an organization that stands out from the pack when it comes to how IT workers are treated? Make your voice heard. Submit your entry now for InformationWeek's People's Choice Award. Full details and a submission form can be found here.