Differences in price between Google's newly available Google Compute Engine service and Amazon Web Services are minimal, but Google's public IaaS offering still comes in at a higher price than Amazon's cloud.
Generally, the market has anticipated a price war
What I'm hearing from customers lately is that the prices are low enough, so they're deprioritizing the sticker price for things like performance and additional features.
Kyle Hilgendorf, an analyst with Gartner Inc.
According to Google's pricing page, the smallest standard on-demand instance of Google Compute Engine (GCE), n1-standard-1-d, comes with one virtual core, 3.75 GB memory and 420 GB local disk for $0.132 per hour.
This is roughly equivalent to Amazon's on-demand M1 Medium instance, which also comes with one virtual core, 3.75 GB memory and 400 GB of storage for $0.120 per hour. Amazon also offers a smaller standard option, the M1 Small, with one virtual core, 1.75 GB memory and 150 GB of storage for $0.06 an hour.
It was this last price that Hewlett-Packard Co. set out to undercut with its OpenStack-based public cloud service launched in December. When Windows Azure Infrastructure as a Service (IaaS) came out last month, Microsoft publicly pledged to match any price drop by Amazon.
It appears Google is taking a different tack.
"What I'm hearing from customers lately is that the prices are low enough, so they're deprioritizing the sticker price for things like performance and additional features," said Kyle Hilgendorf, an analyst with Gartner Inc.
Performance is one area where GCE claims an advantage, offering a private backbone network with less contention -- at least initially -- on the back end. Hard numbers on GCE's performance using industry-standard benchmarks have not been officially published.
SLAs and sub-hourly pricing comparisons
Where this all could come to a head is in the smallest unit of possible purchase -- for AWS, partial instance-hours consumed are billed as full hours, while Google will offer sub-hour billing charges for instances in one-minute increments with a 10-minute minimum.
In terms of availability, Amazon and Google both offer 99.95% service-level agreements (SLAs), but Google calculates availability by month, rather than per year as AWS does, giving its availability guarantee more teeth. Google also pointed to shared-core machines as a means for users to cut costs with its service. These Micro instances start at $0.01 per hour.
The big whammy with GCE, however, is that it still doesn't support Windows. As the more mature service, AWS also has a wider global reach at this point, and more than 30 add-on services that GCE can't necessarily match yet.
Experts expect some of these gaps to close between the two companies. For example, Amazon will probably offer sub-hourly pricing and monthly SLAs at some point soon, predicted David Bartoletti, an analyst with Forrester Research, Inc.
"[Google will] have to prove they can match AWS's history of rapid innovation and new services roll-outs, and of course, compete for mindshare with AWS developers who've had years to become experts in using AWS services," Bartoletti said.
Storage comparison a mixed bag
The largest standard on-demand instance of Google, the n1-standard-8-d, comes with eight virtual cores, 30 GB of memory and 3.54 terabytes of storage for $1.06 per hour. It also offers a diskless option in this size for $0.922 per hour.
Amazon's equivalent standard on-demand instance, the M3 double extra large, also offers eight virtual cores and 30 GB of memory, but users must separately purchase Elastic Block Store (EBS) storage at $0.10 per gigabyte per month, whereas GCE includes storage for this particular instance type in the price.
In general, Google trumpets that its persistent disks support up to 10 terabytes per volume, which it claims is 10 times the industry standard, but if you read the fine print, persistent storage at Google also costs $0.10 per gigabyte per month.
Network pricing comparison
Amazon charges various prices for data transfer in and out depending on the size of the pipe, with a number of options starting with the ability to transfer data into an Amazon region from other AWS services in the same Availability Zone and Region free. It starts charging $0.01 for a data transfer that uses a public or elastic IP address, as well as data transfer from Elastic Compute Cloud (EC2), Amazon Relational Database Service (RDS) and ElastiCache interfaces in another Availability Zone in the same U.S. region.
Data transfer out of Amazon EC2 to the public Internet is another story. Pricing there tops out at $0.12 per gigabyte for 10 terabytes of bandwidth.
For now, Google charges a flat fee of $0.01 for egress (its word for data transfer out) to a different zone in the same region, or to a different region within the U.S. Egress to the Internet goes up to $0.11 for up to 10 TB of bandwidth. All ingress (aka data transfer in) is free.