Google cut its cloud instance prices by 10% this month, putting its low-end IaaS costs just barely below those...
By submitting your personal information, you agree that TechTarget and its partners may contact you regarding relevant content, products and special offers.
of AWS, as the two providers vie for attention and revenues in the cloud computing market.
Prices are just one of the ways Google and Amazon Web Services (AWS) have duked it out in the infrastructure as a service (IaaS) market of late. The two vendors' offerings are similar in service level agreements (SLAs) and support, though AWS still leads in areas such as cloud monitoring tools and the number of data center locations available worldwide.
AWS customers are happy with the frequent pricing cuts Amazon makes.
"We're happy to see the public cloud become a highly competitive market," said Jim O'Neill, CIO for HubSpot Inc., an online marketing firm based in Cambridge, Mass. "All customers benefit when that happens."
As for price cuts, "They come regularly, they come often, they are substantial, and they're always a nice surprise," said Ed Abrams, principal software architect for SynapDx Corp., a biotech research startup based in Lexington, Mass.
That said, it's the flexibility to spin up resources on-demand that keeps Abrams' coming back to AWS.
"When our business pivots and our compute needs change on a dime, I don't have to buy or sell hundreds of thousands of dollars of networking equipment," Abrams said. "I just have to change my settings on AWS and continue."
Still, analysts are skeptical that IaaS pricing will remain a differentiator between the two companies for long.
"I think we're going to see a lot of back and forth between the firms," said James Staten, analyst with Forrester Research Inc. based in Cambridge, Mass. "Everybody cuts prices when they achieve an operational efficiency level that means they can sustain the same margins at a lower price, but eventually each of the major hyperscale vendors will hit the same price-efficiency point, or something similar."
Google's smallest IaaS instance, the f1-micro running a generic Linux OS, is currently priced at $0.012 per hour, while AWS' smallest, cheapest machine -- a t2.micro with one virtual CPU, one GB of memory and running a generic Linux OS in the U.S. East region -- costs $0.013.
But pricing comparisons between the two are trickier when calculating long-term costs. Each vendor offers a certain tier of free services, and discounts for sustained use of resources. Also, while Google advertises its prices in per-hour terms, it bills for the first 10 minutes, then in one-minute increments.
It's also worth noting that while Google's smallest IaaS instance is a fraction lower, AWS offers a much larger variety of instance sizes and prices. Google's largest, most expensive machine, an n1-highmem-16 with 16 virtual cores and 104 gigabytes of memory, running Red Hat Enterprise Linux (RHEL), costs $1.314 per hour, and all RHEL images are charged a one-hour minimum.
Amazon's largest, most expensive machine, meanwhile, is an i2.8xlarge with 32 virtual CPUs, 244 GB of RAM and eight 800 GB solid-state drives, running SQL Standard in the U.S. West (Northern California) region, and is priced at $10.518 per hour. The table below shows how AWS and Google compare on pricing for the smallest and largest instances according to region, OS and cost.
AWS free tier vs. Google quotas
Most of Amazon's free tier is capped at one year, though some free-tier offerings last indefinitely; Google’s free quotas for its Google App Engine platform as a service (PaaS) are each capped at a daily maximum rate and do not extend to its Google Compute Engine infrastructure, though programmers can also get up to 28 instance-hours per month beneath the Google App Engine platform for free. AWS's PaaS, Elastic Beanstalk, is offered at no charge -- users pay for the underlying instances only.
With Amazon's free tier, customers get 750 hours per month of Linux or Windows t2.micro instance usage and 5 GB of Simple Storage Service (S3) storage with 20,000 Get requests and 2,000 Puts requests. Other services with a free tier include Amazon Mobile Analytics, Amazon Cognito, DynamoDB, Relational Database Service, CloudFront, Simple Workflow Service, Simple Queueing Service, Simple Notification Service, Elastic Transcoder, Trusted Advisor and CloudWatch.
Both vendors also offer some network traffic free of charge. Neither charges for data ingress – that is, data entering the cloud provider’s network from outside. Egress to other services and regions within the cloud providers' networks are also free. Google Cloud's Platform Starter Pack and AWS Activate are programs for startups with free hours and resources available.
In some ways comparing services and prices between the two is an apples-to-oranges exercise, according to SynapDx's Abrams, since Google's main focus is PaaS and AWS’ is IaaS.
"If you want to write a Django app in 10 minutes and try it out for free with a hundred of your best friends, Google App Engine lets you do that in minutes, and you can't quite do that in Amazon in that timeframe at all," Abrams said. "But that's not what Amazon is there for; I see them operating in pretty different spaces."
AWS Reserved Instances vs. Google Sustained Use Discounts
As for discounts, AWS customers must sign a one- or three-year contract and pay an upfront fee for Light, Medium or Heavy Utilization Reserved Instances, which are then priced at a discounted hourly rate.
It behooves customers to be exacting about their instance usage when choosing Reserved Instances, as the upfront fee can make Light Utilization instances more expensive than the on-demand price if too heavily utilized. AWS acknowledges this with a chart on its pricing page, which shows the ideal utilization rates for different Reserved Instance types.
While customers can achieve savings of 65% over the on-demand price with Reserved Instances, it can be a challenge to decide whether to use Reserved Instances, particularly when workloads become unpredictable.
"If I buy a Reserved Instance for a certain account and then switch, I have to sell the Reserved Instance on a marketplace," said Fons Van Geelen, IT director at Mirabeau, an Internet service company based in the Netherlands that resells AWS services and is an AWS advanced consulting partner. "You can save a lot of money buying Reserved Instances … but for now it's very hard to make this decision."
Google's approach requires less upfront planning from the customer with Sustained Use Discounts that are applied automatically, depending on monthly usage. Using an instance for more than 25% of a month earns a 20% discount; the next 25% earns another 20%, and so on. With 75% to 100% usage per month, the user pays less than half -- 40% -- of the base rate.
The Sustained Use Discounts from Google may be a pricing differentiator that sticks, Staten said.
"I don't know that AWS would necessarily copy that, and I don't know if Microsoft would either," Staten said. "Both of them certainly could, but it's a different philosophical approach."
AWS vs. Google SLAs, tech support and monitoring
SLAs for the two clouds are similar, though Google's allows for more downtime than AWS. AWS offers an SLA for its Elastic Compute Cloud (EC2) and Elastic Block Store (EBS) that kicks in if two or more availability zones in a particular region are unavailable to the customer at the same time below 99.95% availability per month. If availability is 99.0% to 99.5% for the month, the customer can request 10% of their monthly bill back; at 99.0% and below, they can request a 30% credit.
Customers can request 10% of their monthly bill between 99.00% to 99.95% availability with Google; 25% of their monthly bill between 95.00% and 99.00%; and 50% of their monthly bill for 95.00% or lower, which can mean more than 36 hours of downtime per month. The same SLA applies to Google App Engine; Amazon does not offer an SLA for Elastic Beanstalk. The table below shows SLA terms for each vendor.
Customers can purchase tech support from both vendors at different monthly rates, depending on the type of support needed. Enterprise-level support (called Platinum support in the Google world) starts at $15,000 per month for both companies. Both AWS and Google also offer basic customer service, support forums and documentation free.
As for cloud monitoring, AWS' experience in the market shows, with three generally available products for monitoring usage, costs and even for performing compliance audits of log files in CloudWatch, Trusted Advisor and CloudTrail, respectively. Google purchased a company called Stackdriver in May and has a generally available Google Cloud Monitoring API that programmers can write to, but the Stackdriver IP remains in private beta today.
AWS vs. Google Global locations
AWS' global reach still extends far beyond that of Google. AWS has 10 global regions today, including its GovCloud region for security sensitive government customers, with more than two dozen availability zones distributed within the regions. AWS also has dozens of "points of presence" for its CloudFront Content Delivery Network around the globe, some of which are rumored to become full-fledged AWS data centers in Europe, North America and Southeast Asia.
Google has a total of three regions with eight availability zones, with the majority of them still in the U.S.