Thoughts on technology, movies, and everything else
System administration is a special are of IT. It also has a special place in my heart. It is an interesting mixture of all the other disciplines, both common across the whole industry, and at the same time unique for each person, company, and geographical location. When I have something to say or share about system administration, I use this category.
Faast.js started as a side project to solve the problem of large scale software testing. Serverless functions seemed like a good fit because they could scale up to perform work in parallel, then scale down to eliminate costs when not being used. Even better, all infrastructure would be managed by the cloud provider. It seemed like a dream come true: a giant computer that could be as big as needed for the job at hand, yet could be rented in 100ms increments. But trying to build this on AWS Lambda was challenging: * Complex setup. Lambda throws you into the deep end with IAM roles, permissions, command line tools, web consoles, and special calling conventions. Lambda and other FaaS are oriented towards an event-based processing model, and not optimized for batch processing. * Primitive package dependency support. Everything has to be packaged up manually in a zip file. Every change to the code or tests requires a manual re-deploy. * Native packages. Common testing tools like puppeteer are supported only if they are compiled specially for Lambda. * Persistent infrastructure. Logs, queues, and functions are left behind in the cloud after a job is complete. These incur costs and count towards service limits, so they need to be managed or removed, creating an unnecessary ops burden. * Developer productivity. Debugging, high quality editor support, and other basic productivity tools are awkward or missing from serverless function development tooling. Faast.js was born to solve these and many other practical problems, to make serverless batch processing as simple as possible.
And here’s the quick visualization of the architecture for you.
Komiser is a really nice tool that provides an overview of your Amazon AWS setup. After a super simple install, you’ll have a web console which visualizes your AWS regions and the resources you run in them. It’s great for getting a quick overview, as well as for some analyses of billing, security and utilization issues.
“Packets-per-second limits in EC2” is an interesting dive into network limits on the Amazon EC2. Even if you aren’t hitting any limits yet, this article provides plenty of useful information, including benchmarking tools and quick reference links for Enhanced Networking.
The conclusion of the article is:
By running these experiments, we determined that each EC2 instance type has a packet-per-second budget. Surprisingly, this budget goes toward the total of incoming and outgoing packets. Even more surprisingly, the same budget gets split between multiple network interfaces, with some additional performance penalty. This last result informs against using multiple network interfaces when tuning the system for higher networking performance. The maximum budget for m5.metal and m5.24xlarge is 2.2M packets per second. Given that each HTTP transaction takes at least four packets, we can translate this to a maximum of 550k requests per second on the largest m5 instance with Enhanced Networking enabled.