Off-by-none: Issue #15

December 11, 2018

It's all about Layers...

Welcome to Issue #15 of Off-by-none. I’m glad that you could join us. 😀

Last week we recapped re:Invent and took a look at some of the excellent talks and AWS product releases. This week we’ll dig deeper into Lambda Layers and see how people are having a bit of fun with custom runtimes. Plus we’ve got more talks from re:Invent and plenty of other serverless tidbits for your mental ingestion.

Lots to get to, so let’s get started! 🚄

What to do when AWS gives people access to Custom Lambda Runtimes… ⚙️

AWS already took care of C++ and Rust for us, plus some launch partners have already added PHP and Cobol support as well. But it seems that the community is taking advantage of this new feature in a big way.

The team over at The Agile Monkeys added a Haskell runtime. Think about it, a purely functional programming language running pure functions on stateless serverless functions! Okay, maybe that’s a bit much, but if you’re a hardcore functional programmer, you may want to give this a look. 😎

Graham Krizek added Bash support, which is pretty darn cool. He even included executables like aws, scp, git, wget and a whole lot more. Think about all the interesting and powerful use cases this opens up. Just this git support alone adds a number of possibilities. 🤓

Data scientists rejoice! You can now run R on Lambda thanks to this tutorial by Philipp Schirmer. There might be some memory limitations, but overall this looks like a workable solution for all you number crunchers. 📊

There’s also this proof of concept for a Serverless Open Runtime for AWS Lambda. Definitely an interesting concept, especially the language agnostic middlewares piece. Could turn out to be a terrible idea, but definitely something to keep your eye on. 🤷‍♂️

When you want to know how to use AWS Lambda Layers… 🥞

You can certainly build Lambda Layers on your own, but several companies are now providing them as a way for you to easily instrument your code. Epsagon, PureSec, Thundra, DataDog, IOpipe, and more, have all built Layers that you can simply plug in to your existing Lambda functions without modifying your code. That’s pretty easy.

Of course, our friend Paul Johnston has some thoughts on Lambda Layers and Custom Runtimes, including initial thoughts on best practices.

If you’re looking to help influence the future of Lambda Layers, take a look at this RFC on how to handle permissions with LayerVersions in SAM. AWS always appreciates feedback from the community, so feel free to throw your hat in the ring and add your comments. 🎩

When you refuse to believe you’ve watched all the good re:Invent talks… 📺

Not all of these are available to watch, but there is still a ton of amazing re:Invent content out there that you probably missed, even if you were at re:Invent! Here are three more talks that I found to be super interesting.

Inside AWS: Technology Choices for Modern Applications (SRV305)
Tim Bray, a Senior Principal Engineer from AWS, talks us through how AWS dogfoods serverless to power many of their own services. Even API Gateway runs on Lambda. He notes that “capacity planning sucks” and that you should “use serverless whenever possible.” This talk is full of great advice, including ways to “minimize state hydration”, plus some helpful notes on the three integration patterns. Watch the talk

Reddit’s Serverless & Compute Infrastructure at Scale (STP18)
Anand Mariappan & Jesjit Birak from Reddit take us through their latest redesign process and the steps they took to avoid another incident like “the Digg Mass Exodus of 2010.” The overall process was helpful to understand, but their method for scaling their video ingestion system using serverless tech is really interesting. A great lesson for enterprises here, as they built this to run along side their existing monolith. Watch the talk

Close Loops & Opening Minds: How to Take Control of Systems, Big & Small (ARC337)
Colm MacCárthaigh, another Senior Principal Engineer from AWS, lays out ten patterns to use while building control planes for distributed systems. Since all of our serverless applications are distributed, this makes for a really useful guide when building our own applications. Colm dives a bit into control theory, but keeps the advice practical so that you can apply these techniques immediately. Watch the talk

When you’re still debating what database to use with your serverless app… ⚖️

If you plan on using DynamoDB, you may want to look at Alex DeBrie’s DynamoDB On-Demand: When, why and how to use it in your serverless applications. Plus, lots of your burning DynamoDB questions are answered in here.

If you still want to go the relational database route, check out A crash course on Serverless with AWS — Building APIs with Lambda and Aurora Serverless by Adnan Rahić. This is a great post to get you started, I just wish he didn’t use an MySQL ORM. 🤦🏻‍♂️

And speaking of MySQL, I released a new version of serverless-mysql that fixes an ENQUEUE issue. If you’re not familiar with it, this module helps you with Managing MySQL at Serverless Scale.

What to do when you need more serverless content… 🙏

Jon Vines gives us some ideas about Breaking Down the Serverless Monolith. It’s tempting to load up functions with a lot of capabilities as it keeps things “simple” and is familiar to most developers. Some good lessons learned are outlined in this post.

If you’re interested in learning some more best practices, take a look at Five Essential Principles for Developing Lambdas. I think most of these are pretty solid (especially single-purpose lambdas), plus there are some examples, which is quite helpful.

Another great thing about single-purpose functions is that they can be optimized for their specific job. Case in point, don’t overpay when waiting on remote API calls by using the appropriate memory configurations.

For you serverless security buffs, take a look at Ory Segal’s 6 Cloud Security Predictions for 2019. And if you want some hands-on experience, try going through this Serverless Security Workshop. 🔒

When you remember that Microsoft Azure has serverless functions too… ⚡️

Mikhail Shilkov is Making Sense of Azure Durable Functions for you with his new (very detailed) post. Though the title suggests this is all about Microsoft’s solution, there is quite a bit of background on microservices, event-driven applications, serverless function composition and more. Definitely worth the read if you’ve got 20 minutes or so to spare. 📖

Kate Baroni, a Software Architect at Microsoft Azure, shows us how an Azure Function can orchestrate a real-time, serverless, big data pipeline. Plus, if you’re interested, there are some links to related posts that go into more detail. I love finding interesting use cases like this, but it’s curious to me that Azure is doing complex orchestrations within a single function (with no mention of Durable Functions). This has always been a big anti-pattern with AWS Lambda, but maybe not with Microsoft? 😕

Serverless Star of the Week ⭐️

There is a very long list of people that are doing #ServerlessGood and contributing to the Serverless community. These people deserve recognition for their efforts. So each week, I will mention someone whose recent contribution really stood out to me. I love meeting new people, so if you know someone who deserves recognition, please let me know.

This week’s star is Simon Wardley (@swardley). Simon invented Wardley Maps, which changes the way we look at strategic planning. You can read all about how it works here (and I suggest you do). Beyond that, Simon is a huge proponent of serverless and has been predicting for quite some time that it is the future of computing. He has a number of brilliant talks about serverless (including ServerlessDays Hamburg and Serverlessconf San Francisco 2018), plus his Twitter feed often contains entertaining back-and-forth arguments as to why serverless adoption is inevitable (see this recent Twitter thread). I’m a big fan of Simon and appreciate the work he is doing to make the case for serverless.

Final Thoughts 🤔

Lambda Layers is exposing serverless computing to a number of new communities, and people have been rushing to add support for all kinds of runtimes and service integrations. A recent report by Gartner identified “serverless computing” as the number one key trend for 2019 and noted that “more than 20 percent of global enterprises will have deployed serverless computing technologies by 2020.”

We are still early in this journey, but as Simon Wardley says, “No more questions on serverless. It’s not an ‘if’ but ‘when’. Get on with learning.” This is sage advice, and what we’re encouraging with this newsletter.

I hope you’ve enjoyed this issue of Off-by-none. I love getting your feedback. It is always most welcome and much appreciated. Your suggestions help me make this newsletter better each week. Please feel free to contact me via Twitter, LinkedIn, Facebook, or email and let me know your thoughts, criticisms, and if you’d like to contribute to Off-by-none.

Go build some amazing serverless apps!

Take care,
Jeremy