Videos
Reinvent 2018
video AWS re:Invent 2018: Aurora Serverless: Scalable, Cost-Effective Application Deployment (DAT336)
slides
Slides Aurora Serverless: Scalable, Cost-Effective Application Deployment (DAT336) - AWS re:Invent 2018
AWS Online Tech Talks
video How to Build Highly Scalable Serverless Applications with Amazon Aurora Serverless
Summary
Nice comparison between serverless and cars
Physical Machines ==> Virtual Machines ==> Microservices ==> Serverless
Until you don't need to eliminate capacity from DB you are not really serverless
Key use cases:
unpredictable workloads
cyclical workloads ( development testing weekdays)
intermittent workloads (you don't know when people will access to your data)
If you are a static DB with a higher and continuous transaction this is not for you.
Serverless DB offering
1. Aurora Relations Database (MySQL 5.6 and Postgres now it is in preview)
2. DynamoDB Key-Value
Aurora Serverless
There is a pool to DB different sizes that are there
Keep completely to 0 will cost 30 seconds of waiting, so in some production usage, it is convenient to keep at least a small instance on. (to understand better)
Scales quickly up when the load increase and scale down slowly when traffic decreasing
Rest Data API
It is in preview, simple http interface to connect simple devices (iot, mobiles, watch) to the Aurora database
Pricing:
1 ACU = more or less 2GB memory and CPU
pay per second
$ 0,06 per seconds
DynamoDB
It was already serverless for administration, scaling and lambda integration.
But the scaling was slow and this cause dissatisfaction
With DynamoDB on-demand it scales with milliseconds and it was a good improvement.
The capacity start from 0 and scale also up to max consumption.
Pay per million read and write request