Lambda
AWS announced some very cool improvements to Lambda this year.
More Granular Billing
Lambda is a very affordable AWS service. You pay for the number of times your code is triggered (requests) + the time it takes for your code to execute (duration). The duration of the code was previously being rounded to the nearest 100ms. Amazon announced today that this would be changing. They are now rounding the duration from the nearest 100ms to 1ms. This will give some savings to users with large amounts of traffic and low latency API's.
New for AWS Lambda – 1ms Billing Granularity Adds Cost Savings
Allocation Bumps
A limitation of Lambda Function was the ability to run memory or CPU intensive workloads. You can now take advantage of allocating up to 10GB of memory to a Lambda Function (300% increase). CPU is allocated in proportion to memory so you can now take advantage of up to 6 vCPU's. Paying an uncharge for higher resource allocations may be offset by the lower duration running time. Resulting in savings.
You can see from the following chart that going from 1GB to 5GB costs about the same, but it's 5X faster! Likewise, if you go from 5GB to 10GB you pay an additional ~$20
and the code is 7.6X faster!
Containers
I think this is the most exciting news about Lambda Functions. Lambda is a serverless technology which allows you to run code without provisioning servers, you pay for what you use. Now you can build your Lambda functions on top of docker and they will execute as expected locally. AWS said it best:
"Just like functions packaged as ZIP archives, functions deployed as container images benefit from the same operational simplicity, automatic scaling, high availability, and native integrations with many services."