How to Cut Cloud Costs: Save $1000’s on AWS, GCP, and Azure

Ensure Compute Infrastructure is Sized Correctly

One of the simplest ways to cut costs is to ensure your compute infrastructure is correctly sized these are things like EC2’s or GCE’s servers. For instance, if you’re using an AWS t4g.xlarge instance but only utilising 10% of its CPU, you can downsize to a smaller instance (potentially a t4g.small) and save significantly. Reducing the size of the machine by 75% can result in a thousand dollar saving annually per server.

Example: If an AWS t4g.xlarge instance costs you $0.1344 per hour, downsizing to a t4g.small instance at $0.0168 per hour could save you approximately $1,008 per year for just one instance.

Ensure Database Server Infrastructure is Sized Correctly

Similarly, your database server infrastructure should match your actual usage. If your AWS RDS db.t4g.xlarge instance is running at 10% CPU utilisation (or lower than available memory), downsizing to a smaller instance can drastically reduce costs without compromising performance.

Example: Switching from an AWS RDS db.t4g.xlarge instance at $0.258 per hour to a db.t4g.small instance at $0.032 per hour could save you around $1,952 per year per instance.

Pre-purchase Long Running Infrastructure

For infrastructure that runs continuously, consider pre-purchasing Reserved Instances in AWS or Reservations of Compute Engine on GCP. AWS, GCP, and Azure offer discounts of 30-70% for long-term commitments.

Example: Pre-purchasing a reserved t4g.xlarge instance on AWS for a year can reduce costs from $2,229 per annum to $873.37 if paid all upfront on a 1 year term.

Consider Serverless Solutions

Switching to serverless architectures, such as AWS Lambda or Google Cloud Functions, can lead to significant cost savings. Serverless solutions charge only for actual usage, eliminating the cost of idle resources.

Moving transactional APIs, and sometimes even functionality hosted on Docker or Kubernetes to a Lambda can reduce costs by $100’s each month. AWS charge fractions of a cent of their Lambda Pricing so can be an attractive re-architecting option.

Implement Load Balancing for Efficient Scaling

Load balancing helps manage traffic during peak periods without maintaining large or multiple servers constantly running. This is the equivalent of putting on 10 staff members at McDonalds during the dinner time rush but then reducing to 3 once the dinner time rush is over – but we turn the servers on/off at the times we require them. It ensures resources are used efficiently and scales up or down based on demand and or planned events like sale periods or large EDM send.

Example: Regardless of your server sizes Load Balancing can help distribute traffic evenly, scaling your infrastructure dynamically during high-demand events like sales, thus optimising resource usage and costs. So if you have 5 db.t4g.xlarge servers running costing $185.76 per month each (or $928.80 total for all 5) but the metrics say you only need 1 for 90% of the day and 5 for 10% of the day you can automatically scale the other 4 down for 90% of the day meaning you would save $668 for a total cost now of $260.06.

Utilise Web Application Firewalls (WAF) to Manage Data Transfer Costs

Data transfer costs can escalate quickly, especially with bot traffic abusing your services. Implementing a WAF can mitigate these costs by blocking malicious traffic.

Example: A WAF can prevent bots from consuming large amounts of bandwidth by requesting heavy content like images and videos, thus saving on data transfer costs.

Manage Static IP Costs

AWS now charges $0.005 per hour for static IPs and public IPv4 addresses. This can add an extra $3.60 per month for each server, RDS instance, or other public-facing infrastructure. This might not sound like much but our clients have an average of 10 services using IPv4 in their account which is an easy $432 per annum saving.

Optimise ECS Services and Fargate

While AWS Fargate is excellent for scalability, EC2 instances can be more cost-effective for consistent traffic. Configuring your ECS services to use EC2 instances during predictable, stable traffic periods can lead to significant savings.

Example: Migrating from Fargate to EC2 for steady workloads can lower your costs, as Fargate is priced for flexibility rather than sustained usage.

Optimise Machine Learning Costs with SageMaker

Machine Learning activities are notoriously expensive and optimising costs for machine learning models on AWS on tech such as SageMaker involves setting up auto-scaling for idle periods, choosing the most cost-effective instances (GPU vs. CPU), and ensuring high instance utilisation. As advanced as the technology is when you break it down to the fundamentals it comes back to optimising resource usage. In addition more involved savings plans are available for established models and training runs.

Auditing and optimising your cloud infrastructure can lead to substantial cost savings. Our corporate clients have saved a minimum of 20% on their hosting costs following our audits on their environments up to $18k per month (yes, $18k per month was correct with an oversized database server the main culprit and a little code restructure). If you’re looking to optimise your cloud infrastructure and maximise savings, contact us today to discuss how we can help.