Cloud

I’ve written a couple of scripts to create servers on Hetzner over the years to facilitate small projects and experiementation. My older scripts used the Hetzner CLI which was fine but I wanted a version without that dependency or any other dependencies. So, I rewrote it in bash without the CLI by using curl to call the webservices. Initially, I had used jq to work with the returned JSON but since the requirements were pretty simple, I refactored that out.

Hetzner introduced Ampere Altra Arm64 processors as an option for their cloud servers in April this year. I hadn’t tried them until recently but they seem to perform on par with the Intel offerings at similar price points at least in my anecdotal testing on some small instances. The pricing is quite good (as low as €3.79 per month in the US market). You save €0.50 if you use IPV6 only which might be OK for some use cases.

The UK’s OFCOM has referred AWS and Microsoft to the Competition and Market Authority (CMA) for anticompetitive behavior. The report (pdf) is on point. The major concerns listed are: egress fees technical incompatibility committed spend discounts It’s an interesting read. The cloud (e.g., someone else’s server) is critical for many companies and is becoming increasingly important. It’s a very profitable business for the hyperscalers and deserves some attention by antitrust authorities.

I’ve written about HashiCorp before. They have some very nice infrastructure-as-code (IAC) tools that have many users around the world. These tools used to be open source, but that has now changed. Instead of following the Mozilla Public License v2.0, they are moving to the Business Source License (BUSL). It’s sort of of open sourcish but has made plenty of people mad. I can see the point if you have contributed to a project and a company takes the code away and puts it into a less open license.

I guess online meetings just aren’t good enough. Even Zoom is requiring people to go back to the office. You would think that Zoom would stick to remote work just to eat their own dogfood but I guess not. The bit in the article from an economics professor at Stanford is strange, especially from an economics professor. He says that “If you are paying for office space and high Bay Area salaries it makes sense to operate on a hybrid schedule.

37Signals has a large AWS bill. About $3.2M USD last year. It sounds like they have been proactive at managing the bill and that’s just what it costs to run their business on AWS. They are an interesting and smart company and it seems they are going to try to move some/most of that out of AWS. Dell will be happy to hear that but I don’t know if it will result in significant savings because you now have to hire and manage people to run the show robustly (especially the 8PB of storage).

Sometimes you just dodge a bullet. A little over a year ago, I congratulated Hashicorp on going public and mused a little about buying shares. However, I did not because the valuation seemed a little high. Based on this chart, I would say that’s a correct assessment. I’ve missed on my share of stocks. The one I really blew was Coupang. I fell for that one hook line and sinker. Part of my problem there was that I’ve been to Seoul a couple of times in the past five years or so and really like it.

I use AWS occasionally and on those occasions I almost always discover something new and interesting. Today’s discovery is fairly prosaic but interesting in a number of ways: AWS CloudShell. I’m running a little behind the times here as this has been around for well over a year and I’m somewhat familiar with the GCP and Azure versions of the same basic thing. AWS CloudShell is, as the name says, a browser based shell that is integrated into AWS.

I was glad to see Hashicorp go public the other day. They make some really great tools and I love Terraform. Packer and Vagrant are also very useful tools. It’s nice that an open source company is getting this kind of a splash. I’m thinking about buying some shares but I might wait for a little bit of a pullback. A $15 billion dollar market cap seems steep. I’ll revisit in early 2022 to see what it looks like then.

According to a report in The Register, six of ten servers sold in 2020 went to cloud providers. That’s pretty amazing but also not terribly surprising. We did that transition a few years back when we decommissioned our data center and starting using the cloud exclusively. Running things on the cloud is great but you do have to understand how to make the most of it. Is it really less expensive to run a 24x7 monolithic enterprise application?