Seven months after its last price cut on compute power for the cloud, Google is making good on its promise to drive cloud prices down on Google Compute Engine (GCE) at the rate of Moore’s Law. AWS did not match last year’s price cuts by Google, but instead reduced prices on Reserved Instances this past December. This week's price cut by Google is another strategic move in the Google vs. AWS pricing battle that ultimately benefits cloud consumers.
Google now has a significantly lower cost in standard compute instances vs. AWS (m3) unless you are willing to commit to pay upfront for 3-Year Reserved Instances (RIs). However, committing to a three-year time period for AWS RIs raises the risk that future price cuts will wipe out some of your expected savings if Google continues to reduce prices each year. The picture for high-memory and high-CPU VMs is more mixed, depending on your requirements and purchase options.
The price comparisons below are based on Linux in AWS US-East and Google in the U.S.
For high-memory VMs, we compared cost per GB of RAM in addition to total cost since AWS high-memory instances (r3) offer between 15 and 20 percent more memory. Google offers a lower per-hour price (unless you buy 3-Year RIs from AWS), but when calculated as cost per GB of RAM, AWS gains a price advantage for 1-Year All Upfront RIs and 3-Year RIs.
High-CPU instances for Google and AWS (c3) have the same number of CPU cores, but the processors are different and AWS offers more than double the memory. While AWS specifies the processor used (2.8 GHz Intel Xeon E5-2680 v2), Google notes on its website that the processor could be one of three types (2.6GHz Intel Xeon E5 (Sandy Bridge), 2.5GHz Intel Xeon E5 v2 (Ivy Bridge), or 2.3 GHz Intel Xeon E5 v3 (Haswell) depending on the zone. Hence, it’s difficult to do an apples-to-apples “per unit of compute power” comparison. As a result, we chose the admittedly less-satisfying approach of looking at total hourly costs as well as costs per GB of RAM (given the differences in RAM specs between the two clouds). Google offers a lower per-hour cost (except when compared to AWS 3-Year RIs) but AWS c3 instances offer a lower price per GB of RAM for all purchase options. AWS c4 instance types offer an even faster processor at a nominally higher cost, so these instances may provide an even higher price-performance benefit.
Google’s sustained-use discounts and AWS RIs are not a direct apples-to-apples comparison. AWS RIs require a proactive decision, a contractual commitment, and potentially some upfront payment, while Google’s sustained-use discount is automatically applied. Google has shared its “typical” hourly rates based on the average usage as calculated over all Google Compute Engine users. For those of you with a “do nothing” approach to RIs, Google can offer additional savings over AWS on-demand instances without requiring any action from you. Your individual “typical” per-hour cost will vary depending on the percentage of a month that you are running instances.
To cut your AWS bills, consider purchasing AWS RIs. AWS 1-Year No Upfront RIs can drive costs lower than Google on a per-GB of RAM basis for high-CPU instances, while changing to an upfront payment will drop costs lower than Google for high-memory instances as well.
If you’re willing to go all-in with AWS, choosing 3-Year RIs with the All Upfront payment option can get you lower prices than Google across the board, but continued price cuts by Google will eat into the realized savings.
Over the past nine months, AWS seems to be shifting its focus to differentiate based on features vs. costs. However, it’s been more than a year since we’ve seen significant across-the-board reductions in compute prices from AWS. It would be surprising if AWS didn’t announce price cuts sometime in 2015, but it’s yet to be seen whether the company will try to undercut Google prices or go for a “close enough” strategy. At some point, Google’s price advantages may cause more cloud users to sit up and take notice.
The RightScale 2015 State of the Cloud Report found that most enterprises are implementing multi-cloud strategies and IT teams are seeking to broker a variety of cloud services. The real loser here may be internal data centers, given that Google intends to continue the drive to lower cloud prices.
If you want to analyze the impact of cloud prices on your own cloud spend, get a free trial of RightScale Cloud Analytics to analyze your past cloud usage and create scenarios to forecast future spend with the latest prices on Google, AWS, and other clouds.