With AWS’ EC2 product driving roughly 65% of Amazon’s cloud revenue, it comes as no surprise that when Google launched their public cloud, GCE (Google Compute Engine), or Google’s alternative to EC2 took center stage . As Google has the global infrastructure and deep pockets needed to compete with AWS, GCE will likely become a serious option for AWS customers to consider.However, comparing the two offerings can be challenging both in terms of pricing and performance. In this post we’ll examine pricing comparison for GCE and EC2.
Is it really “Pay for what you use”?
AWS On-Demand pricing model for EC2 allows you to pay as you go. The only catch is that the minimum billing unit is one hour. So even if your instance runs for less than an hour, you still get charged for the full hour. Clearly, you are not just paying for what you use.
On the other hand, Google’s GCE offers by-the-minute billing, with 10 minutes being the minimum billing unit. This pricing model practically ensures that you really pay for what you used.
But the question remains—which cloud is cheaper for users?
Different workloads for different clouds
Google’s by-the-minute billing plays a significant role in GCE’s lower pricing for their on-demand instances. However, not every workload will benefit from GCE’s sub-hour billing and may in fact do better with EC2.
Here at Cloudyn we took a look at how two customer deployments would do on either platform:
Customer A typically runs around 1000 m1.large instances, with ~40 minutes average runtime per instance. As these EMR instances are running with EC2’s On-Demand pricing, GCE’s per minute billing would save the company ~40%.
Customer B’s workload is very similar. His typical job is in the range of 80 minutes, and there are ~800 m1.large instances per job. But he uses AWS’ Spot pricing which is much cheaper than GCE. In his case GCE will actually be ~10% more expensive, despite the sub-hour billing of Google. A typical Spot price for m1.large instance is ~$0.12/hour which is far lower than GCE’s $0.28/instance.
It is clear from the case studies above that for workloads that can use Spot pricing, AWS is significantly cheaper. Additionally, AWS offers Reserved Instance pricing. While this requires you to commit upfront to one year (at a minimum), reserved pricing rates for EC2 are much lower than GCE’s rates.
Are RI and Spot rates enough to keep customers with AWS?
Google’s focus on offering cheaper on-demand prices is a brilliant attack on AWS on two levels. Firstly, Reserved instances and Spot instances do not appeal to new customers who are not fully familiar with the ins and outs of AWS. These customers are not prepared to commit themselves long term to AWS. It could be for this reason that RI dropped by 9.4% in the month of January. Additionally, many current companies view RI costs as CapEx which undermines the primary reason to move to the cloud in the first place. Likewise Spot instances, while super cheap, have significant limitations in that AWS can pull the plug at any moment.
Additionally, Google just might offer their own version of reserved and spot instance pricing, totally eliminating AWS’ edge.
But is it all about cost?
While in terms of pricing GCE seems to be quite competitive, this might not be the whole story.
Issues such as performance, availability, regions served and AWS’ wealth of supporting services and products also play a huge factor in deciding where to deploy.
Have you joined the discussion yet?
Find out more about how GCE stacks up to EC2
We’ll be writing about this issue in the coming weeks, but if you want to compare EC2 with GCE for yourself, make sure to check out Cloudyn’s tools for mapping EC2 and GCE instances based on pricing and computing power.