High Performance Infrastructure (Issue #2): Edge Computing & Distributed Systems

Resource

Blog

Topic

Edge Compute

Date

Aug 09, 2019

Author

A recurring conversation I have with our VP of Developer Relations is about what edge computing really means. And he always helps me arrive at the same conclusion: it means different things for different teams and applications.

In a nutshell, you’re either using the device edge (also referred to as the mobile edge) or infrastructure edge (also referred to as the cloud edge).

We dive into this deeper at the end of this issue, but first, let’s start with some stories about a topic that’s been around for a while: distributed systems.

Distributed Systems

Operating a large distributed system, reliably (Pragmatic Engineer)
A collection of best practices implemented by Uber’s engineering team to ensure the reliability of the company’s massive payments system. Topics covered include monitoring, anomaly detection, alerting, and incident management process.

What is a distributed system? (StackPath)
If you need to brush up on the basics and the different types of distributed systems, we’ve got you covered. Benefits related to horizontal scalability, reliability, and performance are covered—as well as challenges related to scheduling, latency, and observability.

A story of breaking a monolith (Zepworks)
Have a monolithic architecture and find yourself constantly wondering what it would be like to switch over to a microservices architecture—and if it’s worth it? This beautifully architected article is your friend.

Edge Computing

Here are some interesting stories about edge computing for those who already have a firm understanding of the topic:

AI Accelerators at the Edge (Forbes)
To bridge the gap between the public cloud and edge computing, chip manufacturers are building niche, purpose-built accelerators to speed up model inferencing at the edge. Three AI accelerators are covered in this post.

Companies struggle to push workloads to the edge (EE Times)
Three hundred storage professionals and software developers weigh in on the challenges they face with edge compute. The main challenge they face? Having a lack of infrastructure for performing compute directly where data is stored.

Edge Computing?

And here is a breakdown of the article below that reveals just how broad and unstandardized the term “edge computing” is:

Edge Computing in plain English (The Enterprisers Project)
Despite the name of this article, it could leave you scratching your head as there are many different ideas about where compute happens at the edge. However, this article does something quite well: it surfaces different definitions several industry experts have for edge computing.

Here are some of these definitions, from most general to most specific:

  1. “Edge computing is everything not in the cloud”
  2. “Edge computing is the concept of bringing computing services closer to data sources.”
  3. “Edge computing is data analysis that takes place on a device in real-time.”

And here is what each definition is referring to:

  1. The first definition is dangerously general but it’s probably talking about both the device edge and infrastructure edge. The device edge is leveraged by companies like Chic-fil-A that have Kubernetes deployments in 2,000 restaurants. The infrastructure edge is leveraged by anyone using StackPath. We have VMs and containers that are situated at the edge of the Internet. The device edge makes sense for smallers workloads while the infrastructure edge makes sense for larger ones.
  2. The second definition more clearly refers to both the device edge and infrastructure edge in edge computing. The closer you can get to the data source, the better. This definition also foreshadows a potential future where the cloud and edge collide and it becomes hard to separate one from the other. How close is close?
  3. The third definition is clearly talking about the device edge, though it’s going even further than what Chic-fil-A is doing. It’s saying that the edge is where the device does everything itself—it doesn’t even need a Kubernetes deployment on separate hardware. In this case an example of edge computing is a personal computer.

As you can see, there are many different types of edge computing. If you’re interested in edge computing that happens at the infrastructure edge, you can learn more here.

Thanks for reading! These are the best resources and stories we found about the given topics for this issue. If you came across another great piece related to these topics, please tweet it to us with the hashtag #HPI.

Related