Seeing a lot of blog posts about implementing durable execution via writing steps to a a database.
I feel they all miss the point. Persisting steps is the easiest part.
The harder parts are correctness under crashes, network partitions, zombie processes, etc. all while guaranteeing liveliness/progress, low latency, and scalability. Don't get me started on versioning, concurrency, or linking cross-function executions together (distributed RPC).
I feel virtually all articles miss how much harder it gets when you want to provide strong end-to-end properties (I get it, its not obvious) and end up proposing fragile solutions that put the burden of fixing the hard 80% back on the developer.
If I wrote about that, would you read that?
Amazon CloudFront announces cross-account support for VPC origins
Amazon CloudFront announces cross-account support for Virtual Private Cloud (VPC) origins, enabling customers to a... https://t.co/PisjoLabKy
We have our first preview of V4 of the AWS SDK for .NET. Our goal is to modernize the SDK and keep an easy upgrade. We have some good performance changes in this first preview.
https://t.co/H4I08t2bWs
Batch exports are now in public beta.
Built on @temporalio, you can schedule exports of event and persons data to BigQuery, S3, Snowflake, Postgres, and Redshift
Batch exports are:
1. More resilient due retry capabilities (automated & manual) when destinations are unavailable.