Various technical explanations & corrections

Rather than dive bomb into each thread, I’ll sum up here:

What I’m less clear about is if server-less and edge computer are used interchangeably. I thought edge computing was programmable logic within the CDN, and server-less was separate, but sometimes they are used as synonyms.

They are not the same. Edge Computing means that the computer doing the work is one of many that are located in strategic places around the country/globe. The idea is that you get your computing done on a machine that’s closer to you, so you save the network latency running from where you are to some central computer thousands of miles away.

Serverless computing was created as an easier way to do cloud programming. Typically, you have to start up a server process to make calls to it. Then you have to manage the lifetime of that server process. How long it stays up, when additional processes are needed to handle the load, when processes should be shut-down to save resources/money, etc. With serverless, you make an API call and the cloud provider (Amazon, Azure, Cloudflare, Fastly, etc.) starts up the server process for you, then kills the server process when the API call is complete.

The reason all Edge Computing is serverless is because of the many distributed computers used in Edge Computing. It would be really hard to manage a whole bunch of processes on servers distributed around the country/globe, keeping them alive only as needed and adding to them as needed, scaling back as needed, etc. Since serverless removes that headache, that’s the preferred Edge Computing model.

Cloudflare Zero Cold Starts:
First, there’s no problem with tricks like this if in fact they are applicable to some decent percentage of use cases. I don’t need to have my vacation house heat itself super quickly if I can tell it a few hours before I arrive to heat up. However, Cloudflare’s current restriction on the processes living at the top level domain (eg. does make this more of a show trick than an actual benefit in my view.

Again, I’m not real clear on what edge computing has to do with web page loading. Why do you have to load a webpage to do some computing?

“Web page loading” is a colloquialism for making an http/https request (like a “Get” or a “Post”) to have something done. It doesn’t have to return an HTML page to your browser to display. For Edge Computing, I’d guess these are normally HTTPS POST requests, since you’re sending data to the server.

The way Cloudflare’s optimization works is that when you make the serverless API call “HTTPS://…” the first thing that happens is that the “S” (think of that meaning “secure”) has to be handled. That means validating credentials so that you know the server is who it says it is, and maybe (for two-way SSL) the server checking that you are who you say your are. During this credential exchange and validation time Cloudflare will start up the server processes that might be needed to handled the serverless API call.

But, since it only works at the top level domain (, and not children (eg, ) it’s not really practical for most web sites of any decent size, IMO.

As for Edge Computing serverless compared to Amazon’s or Azure’s serverless, there are many things at play beside network latency and price. On the big cloud players you have access to a large number of available services that you don’t have to install or manage like you would on the Edge Computing players. And for types of computing that need access to central data/services, the big cloud players are great. Not everything is edge serviceable. Pros and cons to both approaches.

However, the only problem for Alteryx is that the software licenses are very expensive comparing to other tools. For example, python is the most accessible data analytics open source, which is free.

Python is a computing language, not a data analytics tool like Alteryx. They are not competitors. The thread on hiring software developers to write data analytics for you is pretty much spot on that it’s not only more costly, it’s slower. There was a brag about coming back the next day - with Alteryx you get immediate self-service. My experience is that once people see how useful data is, they take several iterations to hone in on what’s important. Waiting a day (at least) for each iteration is too long in today’s business world.


Rather than dive bomb into each thread, I’ll sum up here

Smorg, that was an incredibly clear and helpful set of explanations. Thanks so much for posting it to the board.


It should be noted that “edge computing” is also used to refer to small embedded chips that may or may not ever connect to a server, for example.

Here is an example organization for this with ~100 companies:


1 Like

It should be noted that “edge computing” is also used to refer to small embedded chips that may or may not ever connect to a server,

Devices have to connect to a network to be considered part of edge computing, as the definition of edge is literally “the edge of the network.” If they don’t connect, then they’re just appliances, like calculators or old-fashioned DVD players (those without streaming capabilities, that is).

That said, the extreme edge of a network is indeed the endpoint device itself. When your iPhone does facial recognition, for instance, it does those calculations internally. That not only protects your biometric data, it also lets you unlock your phone when you don’t have cellular or WiFi service. But, if your phone never ever connects to another device, then it’s not doing edge computing.


Just to clarify…many of the component devices could be used in a product that may or may not ever connect to a network. Or may be connected only intermittently. And the network could be a relatively low bandwidth network such as zigbee, z-wave or BLE (BT low energy).
Actually connecting via one of these wireless networks can actually make these devices cheaper since a low end radio can be less costly than any buttons or other UI.