Cloudscaling Blog

This is coming from the perspective of someone working in platform engineering. The further down the stack you are, the more conservative (aka boring) you have to be, because the blast radius increases. Quoting another great post [0] on this: "The risk profile for infrastructure is different than the applications that run on top of that infrastructure." [0] http://cloudscaling.com/blog/devops/continuous-delusion-at-the-infrast…

No, all good, this is super useful - if you're thinking about this, then it matters to us! Our philosophy is built on the "cattle, not pets" [0] and 12-factor [1] app methodologies. To some extent, the Disco server itself should be treated as disposable. Disco runs your applications, which are just deployments of your code (ie git pulls). There's nothing on the server itself to back up. If a server were to die, you'd spin up a new one, run the install.sh script, and redeploy your apps in about 15 minutes. For application data, our stance is that we believe you should use a dedicated, managed database provider for prod workloads. While we can run a "good enough" postgres as noted, we treat that as a dev/staging tool. Disco handles the stateless application layer, you should entrust your critical stateful data to a service that specializes in that. Finally, re: security, we recommend a fresh Ubuntu 24.04 LTS server, which handles its own OS security updates. Disco only exposes the necessary web and SSH ports, so the attack surface is minimal by default. [0] https://cloudscaling.com/blog/cloud-computing/the-history-of-pets-vs-c… [1] https://12factor.net/

To extend the analogy, I'm reminded of the pets vs cattle debate in the context of managing computer systems, and how that distinction continues to shake out. I suspect that as personal computing moves toward cloud computing, following businesses, we will see tension between local-first AI solutions and cloud AI products. Seeing as how the same megacorps that build personal computer pets and pet products are building cloud computing AI cattle products with AI that gives the benefits of pets with some of the upsides of cattle as well. Local compute vs cloud compute is not the same distinction as that between pets vs cattle, but it has many of the same market participants, and with both Microsoft and Apple making moves to bundle AI capabilities with other cloud products for free with subscription value-add-on features is interesting. These features entice users into the vendor walled gardens, nudging users subtly away from local-first pets. https://cloudscaling.com/blog/cloud-computing/the-history-of-pets-vs-c…

Amazon is expanding their EC2 feature set so rapidly... The pace is mind blowing to me. Last year, Randy Bias estimated EC2 was pulling $220M revenue/yr: http://cloudscaling.com/blog/cloud-computing/amazons-ec2-generating-22… And he estimated an overly conservative 10-20% annual growth. But given the EC2 buzz this year, and personal anecdotes from my friends and colleagues using it, my gut feeling tells me the 2010 revenues will have increased 50-100% over 2009 revenues. Is EC2 profitable to Amazon? Likely very profitable if you want my opinion. It is well accepted in the industry that the dominant cost in large scale datacenters is power and cooling --not hardware, not human resources-- and I keep running numbers in my head and the hourly prices of all instance types are well above power & cooling. Just as an example, we know that this new GPU instance has two 95W Xeon X5570 and two 247W Tesla M2050; assuming (1) a max TDP of 50W for the motherboard and rest of the server, (2) instances are run under full load 100% of the time and always reach these max TDP numbers (unlikely, but follow me for the sake of the argument), (3) Amazon uses servers with 80PLUS power supplies (80% efficient or more), (4) a rather good datacenter with a PUE of 1.3 (power usage effectiveness, which includes overhead from power distribution and cooling; numbers in the range of 1.2-1.4 are often quoted by James Hamilton from the AWS team: http://perspectives.mvdirona.com/), and (5)…

Randy Bias also claims authorship https://cloudscaling.com/blog/cloud-computing/the-history-of-pets-vs-c… this tweet by Tim Bell seems to indicate shared credit with Bill Baker and Randy Bias https://x.com/noggin143/status/354666097691205633 @randybias @dberkholz CERN's presentation of pets and cattle was derived from Randy's (and Bill Baker's previously).

I’m guessing this is a techie reference. Computer servers can be divided into two roles in modern technical architectures… they are either “cattle” or “pets”. Pets are computer servers that have names and are always up. Cattle are servers that are transient, spun up and down as needed and do not necessarily have any long running identity/name as a component of a larger system. http://cloudscaling.com/blog/cloud-computing/the-history-of-pets-vs-ca… The poster is making a critiques of how by treating historical persons contributors as cattle, we are considering them a dispensable, unnamed, a commodity that is easily replaced and interchangeable and missing their very real contribution, value, and humanity.

Probably a reference to the pets v cattle analogy http://cloudscaling.com/blog/cloud-computing/the-history-of-pets-vs-ca…

Pets vs cattle: https://cloudscaling.com/blog/cloud-computing/the-history-of-pets-vs-c…

The author is asking for OpenStack to "Implement transparent APIs from AWS, Azure, GCP for OpenStack" There _was_ an EC2 API that OpenStack nova had back in the day, that became unmaintained and was eventually removed http://cloudscaling.com/blog/openstack/the-future-of-openstacks-ec2-ap… This blog was from 2015, and the end result was that nobody was willing to step up and do the hard work of creating these API compatibility layers. Frankly, there were barely enough resources being put in by large companies, to keep the OpenStack APIs themselves well maintained. So, I think the issue of OpenStack was that everyone suddenly decided to shift to containers, instead of VMs for their deployments, and OpenStack no longer was required.

Why are we still building tools to hand manage VMs in 2021? Am I missing something or is this for raising pets instead of heading cattle[0]? 0. http://cloudscaling.com/blog/cloud-computing/the-history-of-pets-vs-ca…

The original pet v cattle metaphor was indeed inspired by servers, not containers. http://cloudscaling.com/blog/cloud-computing/the-history-of-pets-vs-ca… What makes a given component – server, vm, container, whatever – is not the runtime, but how you deal with it when it gets seriously ill. Pets are taken to the vet or hospital to get treatment. Cattle are.. well, read the article I linked.

Writing data migrations that fail safely is tricky, and requires some extra care. Temp tables, adding new columns instead of modifying, feature flags, etc. all help with this. Basically, you want to run stuff in production with less risk. On the AWS/CloudFormation side, "Cattle, not pets" is the motto[0]. Terraform specifically is really good at capturing the desired vs current state of your infra, and showing you what those changes will be before you apply them. Point being - as long as your automation is captured in Infra-as-code, you shouldn't care too much if something gets destroyed. [0]- http://cloudscaling.com/blog/cloud-computing/the-history-of-pets-vs-ca… [1] - https://www.terraform.io/

"In the old way of doing things, we treat our servers like pets, for example Bob the mail server. If Bob goes down, it’s all hands on deck. The CEO can’t get his email and it’s the end of the world. In the new way, servers are numbered, like cattle in a herd. For example, www001 to www100. When one server goes down, it’s taken out back, shot, and replaced on the line." http://cloudscaling.com/blog/cloud-computing/the-history-of-pets-vs-ca…

More on where "pets vs cattle" comes from: http://cloudscaling.com/blog/cloud-computing/the-history-of-pets-vs-ca…

Pet's vs cattle is an analogy that originated with English-speakers. http://cloudscaling.com/blog/cloud-computing/the-history-of-pets-vs-ca…

I'd suggest looking at ScaleIO from EMC which is free for unsupported use I believe. It is blazing fast, runs on bog standard Linux, and supports K8S and Cinder. It's the most impressive block storage product for high performance I've seen. http://cloudscaling.com/blog/cloud-computing/killing-the-storage-unico…

I'm not totally sold on Bias' viewpoint. He wrote another interesting article declaring the death of hypervisors and the eventual takeover of containers: http://cloudscaling.com/blog/cloud-computing/will-containers-replace-h… When I talked to a guy who had worked on the Xen hypervisor for years, he kept going back to Randy's key requirement for all this to be true: "if configured properly" So this other guy's response was "SELINUX is an armed camp if configured properly, yet we have everyone from major banks to the Pentagon being hacked. Truth is that few people have adequate time to configure security properly in the real world. Something that is "probably" as good as the status quo is a very scary statement for those of us living in the real world."

The actual streaming is from CDN, though: http://cloudscaling.com/blog/cloud-computing/cloud-innovators-netflix-…

C'mon folks...don't reward the blatant blogspam with traffic. The original source is linked right there in the second paragraph: http://cloudscaling.com/blog/cloud-computing/cloud-innovators-netflix-…

I did say that I'd written about it on my site. Here are a couple of links. http://pl.atyp.us/wordpress/?p=2240 http://pl.atyp.us/wordpress/?p=2258 Here's one from Randy Bias, which I had also alluded to. http://cloudscaling.com/blog/cloud-computing/virtual-disks Now, why don't you take Dennis to task for his lack of supporting information? He provides even less information about methodology, and his results are simply not grounded in reality at all. In my experience, application of a double standard strongly implies preference for outcome. Why don't you actualy state your position, and then back it up to the same degree you demand others do? I know this is Kibitzer News and all, but it would be nice to see something besides sniping from the sidelines for once.

The article that this article references was posted earlier today. http://news.ycombinator.com/item?id=855389 actual article: http://cloudscaling.com/blog/cloud-computing/amazons-ec2-generating-22…

Very different experience with AWS/OpenStack/running our own here. We indeed had a largish AWS bill ~50k/month. Decided to move dev/test in house into an OpenStack cluster I built with the help of Mirantis/FuelWeb. Decent experience standing up the cluster but spurts of trouble thereafter. No real win with anything API-like at all with OpenStack. Feels half-baked. Getting help is damn near impossible. Meanwhile in AWS-land, things are humming along. Using more and more of the AWS infrastructure and removing pieces we had previously built in-house. Getting lots of mileage out of boto/Java/awscli and starting to see a real paradigm shift in how to run ephemeral servers instead of the 700 instances 24x7 approach you refer to above. I came from running my own servers, floundered for a bit on AWS, built an OpenStack cluster and strongly prefer AWS for simplicity and sheer flexibility of the API. I'm a year in running on AWS and haven't had a single issue requiring any support. The largest benefit has come from the concepts and projects put out by Netflix. If I didn't see companies like them leading the way, I'd likely be a bigger OpenStack fanboy. See generally: http://www.cloudscaling.com/blog/cloud-computing/openstack-aws/