Good Morning!
It’s officially re:Invent season, or as I say: re:Quinnvent season, and we’re gearing up for it. Don’t miss our Wednesday night drinkup at Atomic Liquors! Last year’s was an absolute blast!
From the Community
Should you use a Lambda Monolith, aka Lambdalith, for your API? Surprisingly I think the answer is frequently "absolutely." I do.
How I Save $0 a Month Hosting Open Source Software in the Cloud is worth the read, because I do something very similar: I do things that aren’t "modern" or "efficient," just because it makes it easy to use those things to learn a whole mess of other stuff. You wouldn’t do this as a company in all likelihood, but you can learn an awful lot by doing things "wrong" in your spare time.
At first I thought that Measuring Latencies Between AWS Availability Zones was going to fall for the trap of "AZ naming is inconsistent between accounts," but nope–keep reading!
The Next Platform has this great primer on the economics of H100 GPU Instance Pricing On AWS. It at least passes the sniff test for me; good stuff here.
Nokia is suing Amazon for some patent stuff. Normally I’d be berating them for it, but y’know? A $1.437T company can fight its own battles.
This walkthrough of using Amazon Bedrock to Write, Schedule, and Post Tweets is a good read. Amazon’s GenAI offerings are cheap-as-free compared to the modern cost of Twitter’s API.
Stop Giving Amazon Tax Breaks For Data Centers. You wouldn’t think this would be controversial; it’s not like they’re going to stop building them, y’know? Plus they come with remarkably few jobs.
Podcasts
Last Week In AWS: AWS Jamming Generative AI Down Our Throats
Last Week In AWS: Check Your Email Security Please
Screaming in the Cloud: How Tech Will Influence the Future of Podcasting with Chris Hill
Choice Cuts
Amazon Athena announces one hour reservations for Provisioned Capacity – Nobody would want or need this if Athena wasn’t at times so painfully slow that you start to wonder if you might not be able to get results sooner via working it out with a pencil yourself.
AWS Neuron adds support for Llama-2 70b model and PyTorch 2.0 – Eight short months after PyTorch launched and people could use it with GPUs that see widespread use, AWS updates their SDK so you can use it with their own Amazon Basics Nvidia H100s.
Filter and Stream Logs from Amazon S3 Logging Buckets into Splunk Using AWS Lambda – It’s rare that I see an AWS technical blog post that’s as factually incorrect as this one is. Let me fix it for you. Step 1: Call your Splunk rep, who has already named their latest yacht after you. Step 2: Tell them exactly what you want them to ingest from the relevant S3 bucket or buckets. Step 3: Let them handle it since you are after all paying them enough money to bootstrap a country’s nuclear program. Should they object, repeat yourself more loudly.
Announcing Amazon EC2 Capacity Blocks for ML to reserve GPU capacity for your machine learning workloads – AWS finds a way to monetize its (admittedly understandable) inability to meet customer demand for GPUs.
GoDaddy benchmarking results in up to 24% better price-performance for their Spark workloads with AWS Graviton2 on Amazon EMR Serverless – AWS has historically been very tightlipped on benchmarks for Graviton vs. Intel, but getting GoDaddy to talk about it on their behalf definitely shoots the elephant in the room.
How contact center leaders can evaluate using generative AI for customer experience – AWS seems to believe that we’ve all been anxiously awaiting their wisdom on how to apply generative AI in the contact center. While they’re busy patting themselves on the back for teaching us how to turn our customer service into a sci-fi movie, perhaps they can also provide a guide on how to salvage human conversations after their bot inevitably asks about the caller’s favorite color three times in a row?
Detect and fix low cardinality indexes in Amazon DocumentDB – Yet a customer using Amazon Basics MongoDB is itself a high cardinality event from what I’ve seen…
How power utilities analyze and detect harmonics issues using power quality and customer usage data with Amazon Timestream: Part 2 – Well ring-a-ding-ding, AWS just came out with a long-winded way of telling us that utility companies can use Amazon Timestream to monitor power quality and customer usage, provided they were insane and also somehow weren’t aware of many better suited tools.
Techniques to improve the state-of-the-art in Cloud FinOps using Amazon Neptune – Using a graph database to optimize your cloud bill is ridiculous. It’s a solution in search of a problem and this ain’t it. May as well use Braket, Amazon’s quantum computing offering; it makes about as much sense.
Mark your calendars for AWS Mainframe Modernization sessions @ re: Invent 2023 – Yes, mark your calendars for re:Invent 1977!
Ready for Flight: Announcing Finch 1.0 GA! – Okay, I like Finch; I’ve been using it for most of the past year as a Docker replacement. I’ve had a couple of weird issues, but to date they’ve all gotten fixed. Two thumbs up on this one.
Aggregating, searching, and visualizing log data from distributed sources with Amazon Athena and Amazon QuickSight – In a move that might be mistaken for ingenuity, AWS has decided to reinvent the wheel (yet again) with Amazon Athena and QuickSight, suggesting you should use them for aggregating, searching, and visualizing data from your logs. Because who needs existing and efficient third-party log management tools when you can get lost in a labyrinth of AWS services, right?
How to monitor Amazon Elastic File System (EFS) storage costs – It’s cute how AWS assumes customers have the time to monitor our EFS storage costs. In other news, AWS tacitly admits that "It’s probably low enough for you not to notice until it’s too late" just isn’t cutting it for budget-conscious customers.
Fullstack generative AI sample app for Amazon Bedrock – This article was taken down after I yelled at them on Twitter for embedding an API key in the demo (and incidentally left in a bunch of placeholder copy for URLs), but given that this is twice in a couple of months this pattern has played out, I’ve included an archived copy of the post. To be very clear: this is a process failure on AWS, not one person (particularly the author! My own drafts are war crimes!) getting it wrong.
Tools
aws-architect.js is a quick way to deploy Node.js microservices to AWS Lambda and API Gateway.
… and that’s what happened Last Week in AWS.
