DynamoDB now has two pricing models:
1. Provisioned, which is cheaper if your load is readily forecasted
and enough load to cover the cost making the forecast.
$0.00074/WCU / 3600 seconds = $0.205 per million writes
2. On-demand, which is more expensive per request, but saves
the effort of capacity planning and the cost of excess
capacity. $1.423 per million Writes
Aurora Global DB
A single Aurora database can now be multi-regional.
• One region is read/write.
• One region is read-only.
• Today, failover is entirely your responsibility, including
RDS on VMWare(preview)
You provide the hardware, AWS provides the software
1. Outsource Database Administration to AWS
2. Burst capacity (scale out) into AWS
3. High Availability and Disaster Recovery via AWS
4. Cloud Migration
(alternative to Database Migration Service)
Managed Service for Kafka (MSK) (public preview)
Feature set very similar to Kinesis - direct competitor
MSK use cases:
• Existing Kafka application code
• Exactly once delivery
• Cheaper if your server is at capacity
• Unit of scaling is more coarse (1 M5.large vs 1 MB/s)
• Scaling is not seamless to application.
• Timeseries Database
• 1000 times faster than RDS for time series data
• Time series specific compression and query engine.
• Built-in interpolation and smoothing
Quantum Ledger Database (preview)
• Blockchain as a service
• No relation to quantum computing
• Immutable and Persistent journal ~= Transaction Log
• Reliable and irrefutable “Created By” and “Updated By”
• Document data model with SQL
• Twice as fast as blockchain => 500 times slower than RDS?
• AWS has been using this internally for billing for “years”
• Finally a means to implement the intent of the US Sarbanes-Oxley
Act of 2002
Lake formation (preview)
• No cost, just co-ordinates other services
• mostly generates AWS Glue (pyspark) modules
• Undifferentiated heavy data lake lifting:
• Tranform to ORC or Parquet
illuminating issues before they become problems