Harish Ganesan CTO of 8KMiles has a very helpful blog, Cloud, Big Data and Mobile, where he shows a nice analytical bent which leads to a lot of practical advice and cost saving tips:
- Use SQS Batch Requests to reduce the number of requests hitting SQS which saves costs. Sending 10 messages in a single batch request which in the example save $30/month.
- Use SQS Long Polling to reduce extra polling requests, cutting down empty receives, which in the example saves ~$600 in empty receive leakage costs.
- Choose the right search technology choice to save costs in AWS by matching your activity pattern to the technology. For a small application with constant load or a heavily utilized search tier or seasonal loads Amazon Cloud Search looks like the cost efficient play.
- Use Amazon CloudFront Price Class to minimize costs by selecting the right Price Class for your audience to potentially reduce delivery costs by excluding Amazon CloudFront’s more expensive edge locations.
- Optimize ElastiCache Cluster costs by right sizing cluster node sizes. For different usage scenarios (heavy, moderate, low) their are optimal instances types. Choosing the right type for the right usage scenario saves money.
- Amazon Auto Scaling can save costs by better matching demand and capacity. Certainly not a new idea but the diagrams, different leakage scenarios (daily spike, weekly fluctuation, seasonal spike), and the explanation of potential savings (substantial) are well done.
- Use Amazon S3 Object Expiration feature to delete old backups, logs, documents, digital media, etc. A leakage of ~20 TB adds up to a tidy ~1650 USD a year.