When Reinvent started with the announcement of EC2s with MacOS, it was an indication that there was a lot more we can expect from the rest of the keynotes. Its been a few hours since Andy Jassy's keynote. I am yet to comprehend to the fact that Andy talked straight for 3 hours with just a few minutes of away time when the other speakers took to the camera. That being said, the key note was packed with announcements. Below are my favorite picks and some learnings.
Only 4% of the workload is on cloud at this moment. This was shocking given the amount of work I see going around me.
I also loved the 8 point philosophy on how to reinvent tech. But we will keep that aside for another post.
Announcements were made in all three categories of compute - instances, containers and lambda. A few that stood out for me -
ECS Anywhere - Run containers using the same simplicity of ECS APIs on-prem. This would help speed up migration to cloud using blue-green / canary deployment methods.
EKS Anywhere - Same like ECS but for managed kubernetes!
- Reduction of billing unit from 100ms to 1ms.
- And the big one, ability to launch lambda's from your custom container images. With the new additional RAM of upto 10GBs and 6vCPUs, this will propel the adoption of Lambda even more.
GP3 EBS Volumes - Your IOPS is now no longer dependent on your EBS size. They did a great job on separating them. Now you can provision custom IOPS irrespective of EBS volume size.
io2 Block Express - 4x throughput wrt existing EBS! This is SAN in cloud.
Coming to my favorite topic -
Aurora Serverless V2 - Aurora will now scale to huge transaction spikes within fraction of second in fine-grained increments. This is a great upside over the last generation which doubled the capacity every time there was a spike. It only comes with MySQL 5.7 compatibility. PostgreSQL compatibility yet to come.
Blabelfish for Amazon Aurora - Now this is a huge one! With this additional layer on top of Aurora PostgreSQL, you can run MS SQL Server with no to very less code changes. Even bigger, this project will be open sourced! You can interact with Blabelfish with T-SQL or the PostgreSQL Interface.
AWS Glue Elastic View - Now you can create materialized views using DynamoDB, Amazon Redshift, Amazon S3, and Amazon Elasticsearch Service, with support for more data stores to follow. This means you can aggregate / collate data from different sources using standard SQL! This will make the life of Data Engineers and Data Scientists quite easy!
Well, that's all for now.