Amazon offers three AWS pricing models that suit the needs of different sized companies at various stages of their life cycles. Their most well known on-demand model charges based on actual usage (hourly compute/storage or I/O capacity consumption). It has created a simple way to reduce costs along with any hesitations that accompany the thought of commitment. AWS also offers the option to optimize cloud costs by reserving resources with Reserved Instances (RI), reducing their price by 30-50%. Finally, for short term or non-mission critical apps, AWS offers Spot Instances, where you negotiate a price and AWS delivers. While the AWS pricing models seem simple to use, if they are given a bit of extra attention with the proper knowledge and planning, you will be able to achieve far greater pricing advantages. Check the following seven guidelines to help you optimize cloud costs with AWS.
1. Remembering the Pareto Principle
According to research we’ve done on our own customer base, we’ve learned that compute and block storage resources (EC2 and EBS) account for over 80% of AWS costs. Adding 10% for S3 and RDS usage then accounts for over 90% of AWS costs. This can help indicate which resources should or should not be considered for price optimization. Check out Cloudyn’s insights for more information.
2. Tracking Reserved Capacity Usage
It’s important to monitor your usage closely and be aware of the capacity that you use. We’ve found that over the past year, customers using AWS didn’t enjoy using RIs because of the fear of being locked in. However, RIs offer a tremendous pricing advantage, reducing costs by 30-60% (and even up to 75% in extreme cases). When it comes to making decisions about RI purchases, having a system in place that tracks usage is key. RIs are a great cost saving option if you plan on using AWS resources for more than six or seven months a year. Learn more about AWS RIs from Cloudyn’s CTO.
3. Small Discount Rate Delta Between Reserved Capacity Payment Options
AWS’ pricing method for RIs offers three payment approaches: All Upfront, No Upfront (paying a fixed monthly amount for RIs), or Partial Upfront (paying part of the cost upfront and the rest in monthly installments). It also reserves capacity allowing you to start resources when needed. With AWS’ new Reserved Instance pricing model, if a reserved resource isn’t in use or created, you’re losing money because AWS will still charge you for the reserved capacity. Additionally, if you have no cash flow constraints, you should use the All Upfront option for most purchases in order to get the largest compound discount.
However, the discounted rates between the All Upfront and Partial Upfront payment options is very small. Depending on your liquidity, you can alternate between All Upfront and Partial Upfront payments without it affecting the discounted rate too much. On the other hand, the discounted rate is substantially lower with the No Upfront option.
4. Repurposing Unused RIs
AWS lets you repurpose RIs, without penalty, if they’re not used for their initial instance configuration. When repurposing RIs that aren’t in use, you can change their size within the same instance family (current or previous generation of the same family), modify their Availability Zones, as long as they’re in the same AWS region, and adjust between EC2-Classic and EC2-VPC. Alternatively, if you have RIs that you are not using, and can’t be repurposed, you can sell them in AWS Marketplace. It’s important to note that AWS allows you to repurpose Linux instances, however, that does not include repurposing from open source Linux to RHEL.
5. Reservations Are Not Only for EC2 Instances
Many websites and blog posts that try to explain AWS pricing, tend to focus on RIs for EC2 instances without mentioning the option of using RIs for relational databases (RDS) and DynamoDB. Databases are a very significant portion of your steady capacity and often run 24/7, making RIs a worthwhile investment.
6. S3, S3 Reduced Redundancy Storage (RRS), and Glacier
AWS offers three types of object storage. The most well known, S3, promises you a certain level of redundancy across availability zones [99.999999999% durability] within a region. If you’re willing to compromise on reduced redundancy, you can purchase Reduced Redundancy Storage (RRS), which costs about 20% less and offers less redundancy [99.99% durability] than S3. Finally, Glacier offers archival services similar to tape backup in traditional data centers. The upside to Glacier is that it’s very cheap, at around one cent ($0.01) per gigabyte per month. The downside is that object retrieval takes around four to five hours.
7. Credits and Discounts
Cloud providers offer discounts as usage increases. As we know, Reserved Instances help reduce costs, but AWS offers an additional discount around the 5-10% range if the total value of your RIs is over $500,000. In some cases, AWS offers credits (such as promotional credits, coupons for attending workshops, credits for charges, etc.) instead of monetary discounts. AWS’ Trusted Advisor can also be used to save on costs for new startups.
Similar to the RI model, AWS offers reduced prices for greater storage usage. For example, AWS S3 charges less as your storage size increases beyond 1/50/500 TB.
As discussed above, the AWS cloud can be a cost effective platform for your services if planned properly. Even if you select the wrong pricing model, AWS allows you to change it, which is the biggest advantage when compared to in-house hosting that leaves you stuck with procured infrastructure and concerns about ROI. The best part of AWS pricing is that it’s constantly decreasing. Amazon has made over 40 cost reductions in the last eight years, and with the increasing competition we might see more price reductions in the near future. Enjoy your cloud!