The Edge, Edge Computing & Fastly

A strict definition of the “Edge” is simply the border line between two networks. The point where Network A connects to Network B is the Edge of a Network and the edge is bridged by a Network Edge device. Pictorially, in a way the non-techie can understand, it looks like this:…

What the following picture simplistically illustrates is what CDN Network Edge devices look like on the internet. The CDN Network Edge devices are represented pictorially as the red dot between the ISP and the internet backbone:…

The Network Edge devices is what Fastly considers the “Edge”. The Network Edge device can also contain what is called a proxy server. What is a Proxy server:

A proxy server acts as a gateway between you and the internet. It’s an intermediary server separating end users from the websites they browse. Proxy servers provide varying levels of functionality, security, and privacy depending on your use case, needs, or company policy.

Why use a Proxy server? Well, a proxy server provides a number of advantages:

  1. Bandwidth savings and improved speeds - When a web browser makes a request for a web page from a device, instead of the data travelling from the Web host/origin server over the internet backbone to the ISP and then to the local Network and then to the device, the data only has to travel from the Network Edge device/Proxy server to the ISP to the Network and then to the device. The Bandwidth savings and improved speeds comes from the data avoiding travel over the Internet backbone. The data only has to transit the ISP. The Proxy server does this through the use of a cache. The proxy server coordinates with the source server to cache documents such as files, images and web pages. When content changes on the source server (web host), the source server (web host) will send a content update over the internet backbone to the Proxy server.

One of Fastly’s desirable features is that they have designed their CDN architecture so that when a company updates their website at the Web Host/Origin server, all the Proxy Servers around the world are updated almost instantly. That sounds like a obviously desirable feature but part of the reason Fastly was born was I believe one of it’s founders was frustrated at some of the legacy CDNs that had slow & inconsistent updating of the cache in locations world wide. What that means in a practical sense is that maybe a proxy server in France might serve up a different and outdated web page than a proxy server in the USA that already had it’s cache updated. Fastly updates cache quickly worldwide and all the caches update at the same time which is a feature that attracted many companies to Fastly

  1. Privacy - One feature of Proxy servers is that they have the ability to hide a internal network from the Internet

  2. Improved security - Companies can do numerous things to improve security like provide a proxy firewall/application firewall/gateway firewall. I am not going to go deeply into this topic because it will make the post longer than it needs to be.

There are more advantages to using a proxy server but I am trying to not write a complete novel in this post.

Now, there has been a lot of discussion on what “Edge Computing” is in various threads that I have participated in recently. While Fastly does operate on the “Edge” of the network, I don’t consider what they do as “Edge Computing”.

Going back to this picture “Edge Computing”… , what I call “Edge Computing” and how I see most people define “Edge Computing” is by placing Compute/Storage right at the “Edge” between Network A and the ISP or Network B and the ISP.

Why do that?

Well, Network A simply connects a bunch of end use devices…which can be computers surfing the internet or AR/VR devices or IoT devices or it can be Amazon’s Alexa or security cameras or autonomous cars or anything that requires a connection to the internet.

If a device, let’s say a Virtual Reality headset needs ultra low latency like 20 ms at most, well data coming all the way from the AWS cloud through the internet backbone, through the ISP then to Network A and then to the device will have far too high latency. If we put compute on the Fastly proxy…well, data must still travel through the ISP and depending on the ISP, the latency still may be too high for AR/VR because maybe the latency across a Comcast (for instance) might be 55 ms.

So, some companies like AT&T and Verizon, for instance, seem to be planning to put a proxy server/compute & storage at the “Edge” between Network A and the ISP where the proxy server/compute & storage connects directly either to a end users device or a end users networking device. Verizon believes they can achieve low enough latency for VR/AR that way:…

AR/VR is just one use case…the end device and use case can be virtually anything. Like I said previously, the end device can be IoT devices or it can be Amazon’s Alexa or security cameras or possibly in the future autonomous cars.

Some reasons to use “Edge Computing” is simply for Latency. Other reasons are to save bandwidth. For instance, Video Surveillance is a Edge Computing use case which is shown pictorially here:…

“Edge Computing” is using video analytics at what is called the cluster in the picture but what I would call the local network level instead of sending that information back through a ISP through a Internet backbone to the cloud. Processing the video locally saves enormous bandwidth as video produces enormous amounts of information and it makes sense to avoid creating internet traffic jams both on the ISP and internet backbone by processing the video locally and only sending relevant information back to the cloud. That makes the whole process more efficient. This is described in this article:

Because of high data volume, compute demands, and latency requirements, cameras are the most challenging ‘things’ in the Internet of Things. Thus, large-scale video analytics could well be edge computing’s ‘killer app.’ Tapping into the potential of recent dramatic increases in the capabilities of computer vision algorithms presents an exciting systems challenge,” the researchers say.

Read More:…

Now the ultimate in Edge computing is to simply place the compute on the device itself, instead of putting the compute at the Local Network level. That has disadvantages with some applications. For instance, putting the compute on a Virtual Reality device makes for a far too bulky a headset and it also makes the device require too large of a battery.

So, especially as 5G evolves, increasingly compute will probably settle in on being placed on the Local Network edge for many devices. I personally believe Edge Computing could evolve to be as large as cloud computing over time.

What has all of this has to do with Fastly? I believe there will still be many use cases in which Fastly’s services will still be desired for many reasons but I am still unsure exactly how disruptive Edge computing might ultimately be for Fastly. Edge Computing might very well not affect Fastly as much as I think.

However, the big money is going to likely be in serving IoT devices on the internet which will far outnumber humans using the internet at some point. Just my opinion but I think a majority of IoT devices/use cases will involve edge computing, meaning compute placed between a local network and the ISP

I still think over time, Fastly should move further out and compete on the Edge between the ISP and the local Network for select customers and select applications as a additional service for select customers. I also believe Fastly Managed CDN services can evolve to do that. Currently, Fastly Managed CDN services looks like this:…

The only thing it takes for Fastly’s Managed CDN service to become a Edge Compute service is to place compute slightly before where the Managed POP is on the diagram which is before the information goes out on a ISP to Fastly’s public POPs. I do not believe that is so far beyond the realm of possibility.