3 ms·
- For consumer compute consider serverless (api gateway and lambda) - this addresses the problem of idle/pre-prod environments. - Data at rest: consider s3, ret
by D3m0lish 9y ago
- For consumer compute consider serverless (api gateway and lambda) - this addresses the problem of idle/pre-prod environments.
- Data at rest: consider s3, retention policies and glacier.
- Consider aggressive retention policies on cloudwatch logs, watch out for unnecessary cloudwatch metrics (I still prefer statsd because of the percentiles)
- If you use Kinesis - monitor In/Out bytes and provision shards as needed, control retention.
- Watch out for cross region/hybrid (on prem) stacks - data transfer is very expensive and not obvious.
- Use reserved instances.
- Remove unused EBS volumes.
- Watch out for high IO EBS volumes.
- Use Cloudfront to cache static content.