This article will be continously updated through re:Invent 2020 (which extends to Jan 2021) - pls bookmark and come back to it for updates!
- Compute related announcements
- Storage updates
- Operations related
- Analytics / Business Intelligence
- Customer Engagement - AWS Connect
- Coming soon - Machine Learning
Following the RSS Feed of AWS "Recent Announcements" - there have been 143 individual announcements since re:Invent 2020 started... Even the slightly filtered "NEWS" feed has 55 items.
Some are "this service is now available in this new region" or "We have 3 new digital training courses for snowcone" or services I do not personally care about (say "AWS Lumberyard" - what even is that!!)
So here is my own personal (biased) take of all the "KEY" announcements with a few lines on what it is. If nothing - this is a one pager just for my own reference with links to the launch announcements.
Intel Mac Mini's in AWS's datacenter available to hire as baremetal for 24 hours at a go... Expensive unless you can justify it...
Most instance type updates are usually faster / cheaper / burstier (on network / storage) - basically slam dunk updates
r5b instance type to run databases or similar RAM/IO heavy apps on EC2
m5zn instance type with "Fastest Intel Xeon Scalable CPU in the Cloud" upto 4.5ghz sustained all-core Turbo boost means one mean/fast cpu!
g4ad instance type with FAST AMD Graphics cards
down from 100ms - your $5 bill for lambda's may have dropped to $3...
give me moar! now more things running in fargate can start to move over..
Storage types typically need careful matching to workload to ensure performance without added costs
The key in the name!
a new type of SSD EBS volume that lets you provision performance independent of storage capacity, and offers a 20% lower price than existing gp2 volume types - basically next gen cheaper / faster storage volumes
Even faster block storage with lower latency - maybe should have been called io3 to avoid confusion. Required reference : EBS Volume Types - since there are 2 HDD and 5 SSD types to choose from..
Support combining single region or cross region replication (previously used to be one or other)
The "underdog" of all announcements not even called out at re:Invent - this is crazy hard to do at scale and was released super quiet.
It's the 2020 version of "I am going to replace you with a small shell script". A new service to help detect and fix "Application" errors via automation which sounds very useful for reducing Ops.
Lots of these new "Insights" services from CloudWatch (Container insights was previous) - better observability of what Lambda's are upto
Key is the "RealTime" bit - pickup immediately on "I am not happy - can I talk to your Manager" for example
For example, when a customer calls a business about a defective washing machine, Amazon Connect Wisdom detects the reference to the washing machine model and presents the agent a suggested article on warranty claim procedures
Link up to Customer profiles that may be in other systems (eg., Salesforce)
Identify customers from their voice - a few banks have been trialling this on their call centres (voice recognition never seems to works for weird accents like mine)
Manage Tasks assigned to call centre agents (eg., call back customer with some update)
AWS's replacement to Docker Hub (50gb free hosting + some other bandwidth limits...)
Many organizations build their own "UI" that developers then use to deploy to AWS or combine some infrastructure as code tool with pipelines to handle their deployments. Proton is AWS's take on making this easier by providing a "framework" to create "Templates" and then manage their deployment and ongoing maintenance.
Why use small zip files when you can create a 10gb container! May help move "data" / Ml models / larger "layers" into Lambda since Lambda' have grown up (see above note on 6vcpu/10GB)
The same underlying distro used by AWS that anyone can run anywhere else (OR AWS could then run for you on-prem or on Azure Or on GCP - just saying!)
No need to manage EC2 instances - just run batch jobs as Fargate tasks instead
More details coming later today...