4 ms·
While S3 was the bulk of it, I wouldn't be surprised if they had used DynamoDB for metadata storage, SQS for job queuing (pushing share data down to clients asy
by phamilton 9y ago
While S3 was the bulk of it, I wouldn't be surprised if they had used DynamoDB for metadata storage, SQS for job queuing (pushing share data down to clients asynchronously), SNS for broadcasting pub/sub, Redshift for data analysis, ASG for deployments, etc.
- scapecast 9y agocost for both DynamoDB and Amazon Redshift can both get out of hands real quickly. I've seen it first hand. With Redshift, the problem is that data teams do not adjust the default configuration of 5 query slots. So when things slow down, instead of increasing concurrency for a given node, they add more nodes. I've seen monthly invoices of $100K+ for a workload that could have been done with less than a $2,000 / month.
- alexnewman 9y agometadata on mysql on bare metal last i checked. it wasn’t on ec2