In both cases, cloud computing cannot be abandoned! ｜ David Hansson

BY DAVID HEINEMEIER HANSSON

Comprehensive Editor: Yu Ting @ TO B New Force, B Impact Host

In David Heinemeier Hansson's Reflections on Leaving Cloud Computing, he specifically mentions two situations where you can't leave cloud computing at the beginning. One is that the flow rate is extremely low; One is complex imbalance. In the era of real large models, once the real AI large model is used, it may be unbalanced, and only by using the public cloud can the benefits be maximized

Dr. Wang Jian once gave a very detailed example in his book. A book translated into Chinese is called "Data Analysis Competition Law", the English name of the book is: Competing on Analytics, not Competing on Analysis, the difference is that Analytics refers to analytics, which is a set of analysis methods that cover the entire process of how data is obtained, how to analyze, and how to use. Taking media as an example, analyzing data is to analyze what kind of perspective effect readers like and improve it, but this is traditional data analysis. Analytics in the online data age is that users are seeing real-time content that matches their preferences in real time. This is online data analysis. The analysis of the former is to reflect the user with data, and the more important thing is to give feedback to the user online. Online is the future. The end game of cloud computing is data online

Of course, we recognize that not all enterprise customers are suitable for cloud computing right now, and that economics and efficiency are paramount. Take, for example, an article by David Heinemeier Hansson. As co-owner and CTO of 37signals, created Basecamp and HEY. Ruby on Rails was created. Author of REWORK, It Doesn't Have to Be Crazy at Work and REMOTE. Racers. Antitrust advocates fight big tech monopolies.

37signals is a privately held web application company that was named one of Wired's top 10 startups to watch in 2008. 37Signals is a well-known name in the web application industry, not only with well-known products such as BaseCamp, Highrise, Backpack, Campfire, etc., but also derived from a classic book of web startups, "Getting Real" and "ReWork" (which became the first on the Amazon bookstore list).

Yu Ting believes that cloud computing is the trend of the future, especially in China, so I have compiled the original translation of this article

A true public cloud is ultimately a service that is affordable enough for everyone to use, and a service that breaks the distrust between Chinese companies and makes all startups and startups willing to use it. Especially small and medium-sized enterprises that are still weak today, but are innovating in some corner.

As a To B self-media, one of my jobs is also to witness the migration of Chinese software companies to the cloud. China's software has not been able to do it, the core is the lack of interaction with users, continuous interaction, basic software to the cloud to have the opportunity to develop products. Compared with the United States, cloud computing has changed China's software industry even more.

The progress of the cloud computing era includes all of us' iterative cognition of public cloud, data online, and data scale.

It includes the evolution of technology and business models of manufacturers.

There are also twists and turns, profit gaps with international vendors, software security issues, reduced remote demand that weakens cloud service usage, customer uncertainty, and more.

It also includes new opportunities in the AI era, such as the formation of a new model as a service model on the basis of IaaS and PaaS services; For example, DingTalk and other intelligence at the level of software interaction.

Each of us is a small splash in the face of technology and the trend of the times. It is both idealistic and skeptical.

David Heinemeier Hansson has antitrust values and works in a position related to "going to the clouds", but he still writes neutrally about the circumstances under which "going to the clouds" cannot be "going down". Therefore, the decision to go to the cloud and go to the cloud should be judged by real enterprise-level customers.

This article is ad-free. Thanks to Notion for the translation

It's about thinking about the future.

Here's the translation:

Basecamp has been running on the cloud for more than a decade, and HEY, which was launched two years ago, has been running in the cloud for over a decade. We've run on Amazon and Google's cloud, as well as on bare metal VMs and Kubernetes. We've seen everything the cloud has to offer and tried most of it. Now it's time to conclude that renting a computer is (for the most part) not cost-effective for a medium-sized company like ours with steady growth. The savings from reduced complexity were never realized. So we're making plans to move away from the cloud.

The cloud excels at two extremes, and we're only relevant to one of them. The first extreme is when your application is very simple and has low traffic, and reducing complexity by using a fully managed service can really save costs. This is the path paved by Heroku and the path followed by other services such as Render. It's an excellent place to start when you don't have customers, and it can still drive your business forward even after you start having some. (Then, once usage spikes and your bill soars to the skyline, you might face a good question, but it's a reasonable trade-off.) ）

The second extreme is when your load is very irregular. When you experience drastic fluctuations or huge spikes in usage. When a baseline is just a fraction of your maximum needs. Or when you don't know if you need ten servers or a hundred. In this case, there's nothing better than the cloud, as we learned when we launched HEY, and suddenly 300,000 users signed up to try our service in three weeks, compared to our forecast of 30,000 users in six months.

But neither of these conditions applies to us today. It never worked for Basecamp. However, by continuing to operate in the cloud, we are paying some of the sometimes almost ridiculously high fees in order to prepare for what may arise. It's like paying a quarter of the value of your home for earthquake insurance in a place where you don't live near a fault. Yes, when an earthquake strikes, if an earthquake outside of two states shakes the ground open, cracking the foundations of your home, you might be happy to have earthquake insurance, but that doesn't feel proportional, right?

Let's take HEY as an example. We pay more than $500,000 annually for database (RDS) and search (ES) services from Amazon. Yes, you do have a lot of data to analyze and store when you're dealing with emails for tens of thousands of customers, but I still think it's pretty ridiculous. Do you know how many powerful servers you can buy with a budget of $500,000 per year?

And the cloud has always argued: of course, you have to manage these machines! It's so much easier in the cloud! The cost savings will be reflected in the labor cost! But that's not the case. Anyone who thinks it's "easy" to run an important service like HEY or Basecamp in the cloud has obviously never tried it. Some things are simpler, others are more complex, but in general, I haven't heard of organizations of our size being able to significantly shrink operations teams simply because they've moved to the cloud.

However, it is indeed an excellent marketing strategy at once. Use "You don't run your own power plant either, do you?" Or, "Are infrastructure services really your core competency?" and so on. Then, with a thick layer of fresh paint, the clouds glow so brightly that only conservatives would consider running their own servers in their shadow.

At the same time, Amazon, in particular, is renting out servers at very high margins. Despite significant investments in future capacity and new services, AWS's profit margin reached almost 30% ($18.5 billion out of $62.2 billion in revenue). Now, "the company says it plans to extend the lifespan of its servers from four to five years and the lifespan of its network equipment from five to six years," that margin is set to skyrocket.

That's no problem! Renting a computer from someone else is definitely expensive. But never in these terms. Advertised as on-demand computing, the cloud sounds futuristic and cool, and isn't as practical as a mundane "rent-a-computer computer," though that's mostly what it is.

But it's not just a matter of cost. It also relates to the type of internet we want to operate in the future. It seems to me that this fragmented miracle now runs mainly on computers owned by a handful of large companies, which is really quite sad. If one of AWS's major regions goes down, it looks like more than half of the internet will go down with it. This is not the goal of DARPA design!

So I think we at 37signals have a responsibility to go against the mainstream. Our business model is very compatible with owning the hardware and amortizing it over many years. Our growth trend is largely predictable. We have professionals who can use their talents like managing an Amazon or Google machine. I think there are a lot of other companies that are in a similar situation.

But before we sail more broadly to lower-cost, decentralized shores, we need to steer the brainstorming away from cloud service marketing about running our own power plants. Until recently, everyone was running their own servers, and the tools that come with the cloud are available on your own machines. Don't let entrenched cloud interests convince you that running your own setup is too complicated. Everyone, even their dogs, did it to enable the internet to set sail, and things have only gotten easier since then.

It's time to dispel the clouds and let the internet shine through. Want to learn more about how, why, and when we're leaving the cloud? Check out the Away from the Cloud episode of the REWORK podcast, where I discussed all of this with our Director of Operations, Eron, and host Kimberly. (Interested Chinese readers can jump to this podcast on their own)

Original link:

https://world.hey.com/dhh/why-we-re-leaving-the-cloud-654b47e0