Happy Tuesday folks; open question here - when was the last time an AWS feature release really excited you? Whether this solved a pain point for you, or opened a new possibility - it’d be interesting to hear. I ask because I feel like it’s been a while since I rushed off to read the docs and try something out, but I remember that being quite a common experience of my first few years on AWS (as in not that I was discovering new stuff, but they were releasing it)
Expanded AWS API integrations for Step Functions. Allowed me to remove a bunch of lambda functions in my workflows.
AWS Security Data Lake… We are super excited about it.
https://aws.amazon.com/blogs/aws/new-visualize-your-vpc-resources-from-amazon-vpc-creation-experience/ this one
Last time I was really excited? Session Manager. This marked the end of having to deal with janky key or cert-based authentication mechanisms for our devs.
Honestly not much has excited me since then.
I had built a system to do hacky cross-account IPAM for our transit gateway and having an out of the box replacement with every feature we could ask for (that is now usable with CDK) was such a relief
Karpenter (open sourced by AWS): https://karpenter.sh/
It opened up a whole new world of EC2 instance type options vs the classic Kubernetes cluster autoscaler and plays nice with spot.
The RDS http endpoint meaning you didn’t have to deal with VPC networking to access your DB securely
Really interesting question . I get your point. I wonder if that sentiment is a sign that we’re hitting some form of asymptote where there’s a rich set of services (aside from creating a managed service for every piece of good OSS frameworks) and that there’s not much more to build at the infrastructure and managed platform level?
Still of all the services, every day I find fundamental limitations. I’d be happy with thoroughness of existing services before adding anything else.
Of course this begs another question. What services should AWS add that don’t exist yet?
> What services should AWS add that don’t exist yet?
I think we could use a modern observability platform with efficient storage and query that can process OpenTelemetry data
CloudWatch and X-Ray are both way behind in terms of search/recall, I would guess due to storage that’s less optimized for ad hoc realtime analysis on arbitrarily wide structured events
Currently AWS also has a bunch of Managed Services that are more difficult to set up and integrate and can cost a lot to use, through ingest, compute, and storage costs that wouldn’t be so high in a bespoke design.
Thanks for your answers folks; I think I was struck that when I joined this Slack I’d regularly watch the channel and be excited about new launches. That channel is busier than ever (lots of services exist now so more updates overall) but definitely it’s less common I see something and think “game changer”
However as has been said, it’s not the case that AWS will just keep innovating at the forefront forever - the more services they have to maintain, more and more niche the tweaks they need to make for their bigger customers.
I think I’d like to see some more bulletproofing added to RDS console operations - I feel like for a managed database system there’s still way to many things I can do casually that can totally hose my production database and are not easily reversible. Blue/green deploys is a start but that implies you also have a whole set of test application architecture.
There’s also more I’d like from IAM, especially around least privilege - I feel very nervous giving my team access to do things in our account but as we grow there’s more need for devs to be able to spin up various services and spin them down again. Some way to say “here’s my sensitive stuff, nobody can touch that, otherwise go nuts” would be nice (and yes I know separate accounts is a thing, but sometimes we need production resources created, but without the ability to delete other production resources). There’s so many permissions involved in a simple operation like creating a packer image, or checking a terrraform plan.
A turnkey solution for streaming dynamoDB data into a SQL database would be amazing - something like glue studio but which already knows your dynamo data formats and can suggest table structures - then the ability to apply that and have it build a read-only database for a given view model
Athena could also get some neat updates - like the ability to preview a given file from a partition before selecting it, then generating a DDL from it - once again without having to go the whole hog of creating a crawler, setting it up, running it, finding it missed something etc
Generally more stuff that responds live and gives me what’s needed now, rather than making me plan it all out first, thus increasing agility in the environment
Here’s a feature I’d like - Cloudfront Pre-warming. We’re using @Edge to resize images dynamically; if we alter the distribution we lose already cached images at the edge meaning subsequent requests are noticably slow. I can easily pre-warm for home locations by visiting pages in a browser, but would be very cool if we could supply a list of URLs that are pre-warmed at all edge locations within the defined price area, at the time of deploy (or at least queued for shortly after)
All good ideas! I’ll leave it at that for the moment since I’m in the middle of something, but did want to acknolwegde this.