Back To Blogs

5 Key FaaS Announcements at Re:Invent 2018

By: Surya Jakhotia

2019-01-22

banner

55000+ attendees with 100+ products and services announcements make AWS re:Invent 2018 one of the largest tech events. It was a big event for serverless in general where AWS Lambda continued to be the top trending service from the event.


Here are 5 key releases/upgrades in the FaaS world along with my notes of when to use it and when not to:

1. Invoke Lambda from Application Load Balancer (ALB): You can now expose your Lambda Functions as an endpoint through ALB. This is an alternative to AWS API Gateway where your Lambda can be triggered when a request matches a given host header/path.

When to use it?

· Organizations that are migrating their workloads or have a mix of (micro)services that operate in serverless and non-serverless services will greatly benefit from a simple ALB setup.

· If you were considering an alternative to AWS API Gateway be it for cost and/or setup and execution overhead, and/or their 30 second limit, you now have options.

When not to use it?

· If you were looking to migrate from API Gateway to ALB only for cost, consider that you might need to setup a ALB health check, which in this case will constantly ping your Lambda. Depending on your use case, you might be able to get away without setting up a health check though.

· If your endpoint requires authentication/authorization/caching/throttling then go with API Gateway.

2. IDE Integration: AWS toolkit now enables deeper integration in your favorite IDE. The toolkit has not been released to the VS Code marketplace, so in order to try it you must build and run from source.

When to use it?

· Invoke your Lambda’s w/o leaving your IDE.

vscode
AWS Toolkit Integration in VS Code

lambda
Execute Lambda in VS Code

When not to use it?

· “Deploy Lambda Function” isn’t functional yet. Also what you get is a list of Lambda’s and if your account has 100’s of them, it is painful to scroll down to “your Lambda”. To be fair though, this is an OSS project so expect things to change fast.

3. Lambda Layers: Layers is a packaging solution for your Lambda dependencies including common libraries and/or custom runtime. Your Lambda can reference Layers created by you or use public* Layers* published by AWS and other AWS customers.

When to use it?

· If you want to save on your deployment packages size and/or bandwidth.

· Package your runtime as Layers.

When not to use it?

· If your runtime has a package manager, having your dependencies deployed independently causes more harm than good. Your CI/CD pipeline, unit tests, local debugging will now require pulling these packages (*Layers) *that are outside your package manager. If you already have a CI/CD pipeline for deploying Lambda’s you might as well as skip this announcement.

4. Bring Your Own Runtime (BYOR): You now have the ability to bring your own runtime to execute your Lambda function.

When to use it?

· Author your Lambda functionality in C++, Rust or bash.

When not to use it?

· BYOR means someone from your organization is taking the responsibility of managing the runtime and that includes updates/patching/versioning/testing/deploying your runtime on Amazon Linux image, which in some ways is antithesis to the whole serverless paradigm.

. Ruby 2.5 is now available as a runtime for your Lambda, so you can use it without packaging your own layer.

5. Firecracker: Firecracker is a virtualization technology that is purpose-built for creating and managing AWS Lambda’s and this is now open source. Firecracker allows you to create minimalist-designed micro VMs with boot time of 125ms, which blurs the lines between VMs and containers. You can spin up 4000 of them one machine, which showcases how AWS scales your Lambda’s.

When to use it?

· It is too early to say — If this becomes more of an industry standard to run FaaS workloads, more scenarios become possible (vendor lock-in becomes less of a concern, run mixed workloads helping with transitions between on-prem to cloud), so stay tuned.

Note: The current OSS version runs on bare metal on AWS (nested virtualization is disabled) and on Intel architecture only.

When not to use it?

· Not ready to run your production workloads, yet! Check out this talk: AWS Lambda under the hood to learn about the complexity of operating your AWS Lambda’s at scale.

· It is built on top of KVM, so you cannot develop or run it on macOS natively (VirtualBox/VMware Fusion is your option).


At Re:Invent 2018 we saw customers sharing their use cases of how they are successfully running production workloads using a healthy mix of serverless services. There were also a breadth of announcements making serverless services more deeply integrated with the rest of the AWS ecosystem. ALB Integration, BYOR, Lambda Layers and IDE support makes FaaS and in general serverless go more mainstream than ever.

Originally published at https://medium.com/@suryaj/analysis-of-5-key-faas-announcements-at-re-invent-2018-f73b417b83f0.

comments powered by Disqus