I am sitting in SFO heading home from the AWS San Francisco Summit. Here are the big take aways as I see them…
The Big Push
Looking at the agenda it was abundantly clear that AWS is pushing serverless, containers, and AI. At this point using EC2 instances almost seems like an anti-pattern!
There seems to be a new market where companies are providing end to end managed services for container build, integration, and deployment. Just check your code into Git the kick off a UI based CI workflow. Pretty cool stuff, though I’m not sure how long that market will exist. In my opinion, docker is already pretty easy to use… but I suppose that having a uniform CI pipeline for all of your apps isn’t a bad idea.
Lambda still seems to be on the periphery. I think that most people haven’t fully caught on to this being the new world yet.
With regards to AI, I saw some cool stuff in the maker section of the Summit. One project used Rekognition to identify faces in video. Another was a prototype for an Alexa to Alexa social messaging system. My favorite project had a Nerf gun hooked up to facial recognition which only shoots people it doesn’t recognize. It was programmed to recognize Bezos… but Werner wasn’t so lucky!
A demo of the new Alexa skill builder was a bit rocky, but live demos are a crapshoot any time. The reality is that building interactive voice apps can be done with little coding. We absolutely live in the future.
Redshift Spectrum
This was the big announcement of the event. Redshift now has the ability to reach out and query S3 directly, then join the results to tables that exist in Redshift.
Redshift Spectrum supports a subset of Athena functionality. A notable gap is that Spectrum does not support querying of JSON files, while Athena does. This makes sense from a Redshift perspective, though I would like to see flattened JSON supported.
In my opinion, the way that Spectrum wants to work is as follows:
From our perspective the ability to query S3 is a smart move against Redshift’s biggest competition at this time, Snowflake.
If you want the tech details on Spectrum, check out my other blog where I dive in.
DynamoDB Accelerator (DAX)
The story here is that DynamoDB, known for it’s consistent, single millisecond query times, often required in-memory caching for applications where 4ms is too slow. While there are many options in this area, they require developers to use different API when interacting with the cache.
DAX promises the ability to have query times in microseconds, while being API compatible with DynamoDB. I can see this being quite useful for customers with incredibly high volume website backends in DynamoDB, though it goes slightly against the grain of the serverless ideal. Still… smart of them to simplify the coding side of the equation.
One Cloud to Smash Them All
The keynote started with Werner showing us how AWS is hammering the competition. Microsoft and Oracle are (apparently) slowing down in their growth, while Amazon is poised to take over the world.
Conclusion
I am on board with the serverless and container revolution. I don’t see much reason to use EC2 at all anymore unless you are running a database like Vertica, or doing something freaky with GPU, FPGA, or you have some custom codebase that needs all the cycles. The AWS services where instances are used (Redshift, RDS, Elasticache, etc) all operate as managed services where you don’t need to connect to, or manage the instances.
This makes me smile.
