AWS re:Invent 2023 is nearing the top. This yr’s Keynote by Dr. Werner Vogels as ordinary, didn’t disappoint, you’ll see why in a minute however actual fast earlier than we get into it: In case you are on the lookout for a number of thrilling new bulletins this isn’t the one to observe.
After a, now conventional, “The Matrix” introduction and total theme, Werner went into the subject of value administration and he went deep! I extremely suggest this keynote to these old-school IT professionals with software program growth or knowledge heart administration expertise. You’re in for a deal with!
Alright, let’s get into the main points:
At this level I didn’t know your entire presentation was going to be centered round cost-management within the cloud however I used to be intrigued by the e-book “The Frugal Architect” that he stored referring to, which is a e-book about designing purposes that use sources effectively to avoid wasting computing energy, reminiscence and in flip: operational bills. A fast Amazon search revealed: such a e-book doesn’t exist. Extra on that later.
As soon as it was clear that his total presentation was going to be round this subject, it began falling into place. He began to hit particular factors after which expanded on these. Right here’s a style.
Align value to Enterprise
I actually cherished this level. Within the AWS world, we will get tremendous enthusiastic about options: high-availability, auto scaling and serverless.
However we should always always remember that if our firm’s revenue is determined by low-cost computing, then maybe we shouldn’t be beneath using a super-expensive 4xlarge EC2 occasion if we may very well be doing the identical job with a bunch of smaller, spot situations.
This might not be evident at first, however because the enterprise grows you actually don’t need surprises by way of bills that immediately have an effect on the corporate’s income.
That is one thing that I already do , as a result of my Software program Growth background: Maintain prices in thoughts and by ‘prices’, I imply every thing: CPU cycles, storage, variety of servers, and so forth.
I agree with Werner that Amazon Internet Providers is an incredible service for all of your computing wants, simply don’t let that month-to-month invoice run away from you by accepting defaults or losing sources.
Observability
One in every of his factors was that an software that isn’t tracked and measured will incur in hidden or surprising prices and this level was a pleasant segway to introduce CloudWatch Functions alerts, a brand new function to trace application-specific value and utilization.
Languages
At one level, he was very particular about programming languages and their total footprint and affect within the velocity of our code. Quicker, extra environment friendly languages result in higher code that may get the job executed quicker. He went so far as saying we must be coding in Rust. This is because of its effectivity and velocity. I may argue towards this:
Granted, Python, Java and .NET Languages are fairly heavy as a result of their underlying help platform — making them pointless for brief, transactional applications. However, he did not account for Growth Prices, long-term upkeep and Time-to-Market. Discovering Python and Java builders is sort of easy as these are well-liked languages everywhere in the world. Discovering Rust builders? not so certain about this one.
In fact, if we shift our focus again to his level: Operational value.
A program in Rust, C or C++ that may run in a 100 milliseconds will all the time outperform the identical program written in Python, Java or C# merely due to the super-long load time of the surroundings itself. So, he’s 100% right by way of value financial savings and sustainability.
He additionally touched on the phrase “however, we’ve all the time executed issues this fashion…”, attempting to say that we shouldn’t be afraid of a brand new programming language or expertise to get the job executed in a way more environment friendly and sustainable method. Whereas I agree with this, not all companies can afford to rework their Senior Python builders into Junior Rust builders whereas anticipating the identical stage of output from them, so, your mileage might range!
Gen AI
Once we bought to this a part of the dialog, I assumed “Oh boy, right here we go!” and I used to be anticipating the dialog to tangent wildly into language fashions, picture technology, Amazon Q and so forth, however no! It was the exact opposite of what I had in thoughts.
As an alternative, he confirmed us use-cases of conventional AI (Machine Studying, SageMaker, Imaginative and prescient) to resolve real-world issues, resembling deciphering radiology scans, accurately figuring out grains of rice for germination and analyzing picture knowledge to seek out and assist victims of kid abuse.
By the way in which, about that software program that checks these x-rays photos, Dr Vogels has a background within the well being business earlier than making the transfer to expertise, so, he wrote the preliminary code himself utilizing Python earlier than it was delegated. This code is now open supply and far more feature-rich.
Even on this a part of the dialog he stayed conventional versus leaping within the bandwagon of Generative AI. I adore it!
Though, not gonna lie: I’m an enormous advocate of utilizing the Cloud Growth Package and he occurred to say that there are new constructs obtainable, particular to GenAI to assist us shortly deploy these options for our personal, customized wants.
AI predicts, people resolve
He additionally emphasised that “AI Predicts, however in the end people make the choices”, implying that machines aren’t going to take our jobs, change our medical doctors or develop meals for us, however they’ll actually help us to assist sustain with an ever-growing inhabitants.
As a part of his closing argument, he recommends studying his brief e-book, The Frugal Architect to assist us keep in mind the details of his dialog.
To wrap this up: It was nice! It was actually geared at old-timers from the very begin. In actual fact, within the first minute he checked out a display screen and mentioned “is {that a} PERL script?”, I couldn’t assist however snigger out loud at this one.
Even after the shut it was nonetheless hilarious: “Hey Werner, Can I scan my container builds for vulnerabilities in my CI/CD pipeline?” “Now you can!” — good option to sneak in yet another new function which I’ll actually look into straight away, since I’m a DevOps man.
Now, go construct one thing!
Helpful sources from this presentation
In the long run he casually dropped this e-book that he wrote, which summarizes the identical bullet factors that he hit throughout the presentation. this data is nice no matter cloud computing or not. So, even if you happen to’re not within the cloud but, you must test it out.
By the way in which, it’s a actually brief learn, so, I extremely suggest you’re taking a couple of minutes of your time and go test it out proper now!
CDK
That is the brand new set of constructs that I discussed, if you’re in want of deploying customized, generative AI options in a rush, you must severely check out this: https://github.com/awslabs/generative-ai-cdk-constructs