Denys Rudyi - Fotolia


Taking the mystery out of Google Compute Engine pricing

As the cloud market's IaaS war ensues, pricing battles between major cloud providers heat up. Will Google Compute Engine fit in your cloud budget?

When choosing a cloud provider, cost is an obvious factor. And, in such a highly competitive market, cloud provider...

pricing models can vary drastically. Google Compute Engine pricing, for instance, is different than Amazon Web Services (AWS). For many cloud users, this is a good thing. Here are some key details to keep in mind, as you assess Google Compute Engine costs.

Google Compute Engine, Microsoft Azure and AWS each charge their customers based on cloud resource usage. Servers with higher capacity and performance always cost more than those with lower capacity and performance. The more you store, the more you pay for storage. It sounds simple. But to get the full picture, it's important to read every service-level agreement detail.

One difference between Google and AWS is immediately apparent in the pricing descriptions. While AWS bills by the hour, Google has a 10-minute minimum and then users pay per minute. For example, if you use Compute Engine for 30 minutes, Google charges for 30 minutes. In the same scenario, AWS would charge for a full hour.

Google offers several machine types, including standard, shared-core, high memory and high CPU instances. Standard instances range from one core at $0.049 per hour to 32 cores at $1.538 per hour with 3.75 GB to 120 GB of memory. Shared-core machines only come in two types: f1-micro and g1-small. While the micro instance has .60 GB of memory and typically runs at $0.010 per hour, the small instance utilizes 1.70 GB of memory at $0.025 per hour. Both instance types offer one virtual core.

The high memory machine types, which are usually the most expensive, average between $0.113 and $1.806 per hour. High memory instances offer two to 32 cores, and range from 13 GB and 208 GB of memory. In comparison, high CPU instances also offer two to 32 cores, but include 1.8 GB to 28.8 GB of memory. High CPU machines cost $0.061 to $0.976 per hour.

Bargain hunting with GCE discounts

Although this pricing scheme seems straightforward, automatic discounts based on sustained usage can complicate matters. Depending on how long an instance runs during a billing cycle, Google Compute Engine applies a different discount per machine. If an instance runs for 25% or less of the billing cycle, there is no discount. But if a machine runs for more than 75% of the billing cycle, discounts can reach 60%. Google calculates sustained usage discounts at the end of the monthly billing cycle and applies the credit to the user's account.

To qualify for sustained usage, run a single instance for an extended period of time. In some cases, using multiple of the same instance types at different times also results in a discount. These inferred instances combine "multiple, non-overlapping instances of the same instance type in the same zone into a single instance for billing," according to Google.

Google also charges for using non-open source operating systems, such as Windows, Red Hat Enterprise Linux (RHEL) and SUSE. Billing for each premium OS is slightly different. For example, RHEL images cost $0.06 per hour for instances with less than eight cores and $0.13 per hour for all others. On the other hand, Windows server images cost $0.02 per hour for shared core instances and $0.04 for each core per hour for all other instance types. Additionally, all RHEL images are charged a one-hour minimum, while Windows images are charged a 10-minute minimum. Google offers a complete list of premium OS pricing.

Without tools to calculate all eligible discounts, it's difficult to determine server charges. Fortunately, other charges are more straightforward. Google's network egress charges range from $0.12 per GB for under one TB of traffic per month to $0.08 per GB for over 10 TB of traffic. Load balancing costs $0.008 per GB of data processed, as well as $0.025 for the first five balancing rules and $0.010 for each rule after that.

Virtual private network traffic is charged per tunnel at $0.05 per hour. For persistent storage, charges are based on disk type. Storage on standard provisioned disks runs $0.04 per GB per month. SSD storage costs $0.17 per GB per month. And snapshot storage charges $0.026 per GB per month. While unused static IP addresses cost $0.01 per hour, IPs in-use carry no charge.

There are many components to keep track of when estimating total costs. To simplify the process, Google offers a fairly comprehensive calculator. The pricing calculator allows users to input usage estimates for more than Compute Engine usage. Users can also estimate costs for App Engine, Datastore, BigQuery and Cloud DNS.

About the author:
Dan Sullivan holds a master of science degree and is an author, systems architect and consultant with more than 20 years of IT experience. He has had engagements in advanced analytics, systems architecture, database design, enterprise security and business intelligence. He has worked in a broad range of industries, including financial services, manufacturing, pharmaceuticals, software development, government, retail and education. Dan has written extensively about topics that range from data warehousing, cloud computing and advanced analytics to security management, collaboration and text mining.

Next Steps

Comparing Google Compute Engine and App Engine

Predicting the future for Google Compute Engine

Google adds Windows, Exchange support to GCE

Sizing up AWS vs. Azure vs. Google Compute Engine

Dig Deeper on Cloud pricing and cost optimization