Continuous Delivery Foundation GitOps Summit 2021, 9 Jul 2021

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: GitOps for Great Apes: Policy, PaaS, or YOLO? - Rosalind Benoit, Themist

Description

GitOps for Great Apes: Policy, PaaS, or YOLO? - Rosalind Benoit, Themist

This talk takes a look at GitOps and Policy Automation through the lens of our cognitive roots. We'll look at software tool use and how it helps us scale the thinking processes we share with other tailless primates. We'll address what Policy solves for GitOps and whether or not our current implementation supports this. In the year of human-centered tech, we'll take a look at cognitive neuroscience, what it means to be a fundamentally human engineer, and how tools can adapt. Check out a demo of a new platform and understand why the problems it aims to solve, matter to your fellow humans.

For more Continuous Delivery Foundation content, check out our blog: https://cd.foundation/blog/

A

Hi, I am rosalind benoit and delighted to be here with you today at cdcon to talk a little bit about get ops and the yolo mentality, so I'll go ahead and share. My presentation here get ops has been around for a while, but we saw the term gather some marketing steam a lot during 2020, which makes sense as we had entered the yolo economy and we're all trying to live our best lives right now. A question I spend a lot of time.

A

Thinking about is: how can we use policy with git ops so that we can collectively live our best lives so far outside of the tech giants, who have massive custom tooling, like borg driving their internal development platforms policy, seems like the best option for making it possible for us to do iterative, creative yolo style engineering without regrets about protecting the company's interests.

A

Larger enterprises are using custom, automation, more and more for developer platforms, but it's a huge investment and it's still a challenge to set up guardrails and maintain them. I worked with the spinnaker community for a few years as they adapted custom, automation, built and optimized for netflix workloads for their own needs, and that brought its own challenges.

A

But today I want to walk you through my journey through the yolo economy so far, and in so much as it allows me to share reflections on how get ops intersects with neuroscience in primates and what we can learn from that, and I also want to demo what I've landed on.

A

I started a company this year with some wonderful women of devops. Our goal was to build something that would allow engineers to be engineers and share context, while doing so so that the company could boost roi from people. Investment move faster, of course, build better software without reinventing the wheel over it.

A

Part of my job was to sell folks on this concept of context, and I really believe in it if I think of the way that we develop software like cooking, we start with this recipe, that's kind of known and we customize it. We improvise we have our own special optimizations that come from the learnings that we picked up throughout our careers.

A

I told this story about my grandmother how I have all of her recipe cards with her handwritten notes in the margins and when I'm in the kitchen. That gets me about eighty percent of the way there, but it's that twenty percent that she never wrote down, that really made everything taste like home and that's what I can never get back since she died a few years.

A

So with this story, I need to bring home how much value there is in what we learn from experience, how difficult it is to replace that knowledge and how much risk and investments tied up in that at enterprises today, as they try to compete with software at feminist. We built this product around context that would automatically capture context all the way through engineering, workflows and store that context, so that everyone working in the software life cycle could use it. We thought a lot about how to capture the.

A

Why, when engineers do make decisions, and we were building a way to get that context automatically, instead of expecting folks to put it in their readme files or socialize everything that they're doing or update a spreadsheet or whatever folks are doing right now to keep track of context. We wanted to automate all of that now most of the complex y's that we need to keep track of are around how we use and combine our tools.

A

We wanted to abstract away a lot of that complexity for our users, but keep the freedom to use whatever tool they wanted to use, and we ended up creating this environment that folks needed to work inside of so that we could apply policies directly, while gathering data from the work and mapping everything on the back end as context our tagline was themis makes it possible to put context everywhere whether people are using using git ops, using configuration management to wire everything up or their team's legacy process.

A

So this special qualifier was key, putting context everywhere, no matter what tools you're using all right, no matter what tools you're using first, I want to take a pause and introduce great apes. These are orangutans gorillas, pans, chimpanzees and bonobos. um Those are pans and uh maybe rosalind, of course, humans. We are the tailless primates and we're all really closely related.

A

We all use tools and we've known that uh since 2007 we've known that chimps use stone tools before that, we thought that our fellow great apes only really used the simple plant-based tools like sticks and stuff, but what's been discovered, is that primate tool ingenuity is really driven based on the likelihood of encountering the tool materials. So folks who live in trees have tree parts to work with, and it's about, what's available, plus having the attractive resources that incentivize using a tool, a beehive full of honey.

A

This validates what we learn at work, stress and scarcity- aren't what drive progress nearly as much as having an incentive and the resources to achieve that incentive. So necessity is not actually the mother of invention opportunity at feminist.

A

We came up with the no matter which tools you're using tenant in part, because we know that the one tool to rule them all approach that we see sometimes in our space, is a fallacy, and also when we were doing user research for this and for several other cd tools that I've worked on most managers and executives say we're not going to buy anything that makes our developers switch or consolidate their tools.

A

But since I'm currently a free agent, I want to ask the question: does enabling companies not to be opinionated about their tools really help them? I'd say it depends in terms of git ops. There are specific tools that we really need people to be using like kubernetes, so that's an opinion right there.

A

This is one I think we can all get behind, at least until something better comes along. Really, I've learned if you're not opinionated enough. A lot of time gets wasted on churn. It's hard to capture value from your investment in the longer term, and human investment becomes this weak link to maintaining workloads and processes where we have to be careful with tools and software is. How can we be opinionated in the right way so that having freedom is empowering and not overwhelming?

A

I think having very challenging combinations of tools or too many tools can create stress, and it can even create a scarcity mindset, because we competitive we're worried that we can't keep up that will be outperformed out maneuvered or maybe we're just low resourced, because there's not enough time in the day built in for getting up to speed with the new chain.

A

All this shuts down ingenuity, because it activates fear and the limbics. If you want to learn more about that, you can read my posts from last year want to know how to keep the smartest parts of your brain online as much as possible learn about the limbic system. I won't blabber about it today too much so I started asking this question about this, though also about git ops.

A

Am I happy with this because it's what I know or because it solves the problem and when we're building tools for enterprise software, I think asking these hard questions should drive our ingenuity.

A

If the key to the way that we use tools is that we work with what we have in our environment, then we have to think hard about what are the raw materials we can provide. That will enable our users to make their ideas a reality and collaborate effectively, while measuring up to the standards and constraints that are important to their customers and to the company.

A

This is why user research and qualitative insight gathering is super important, because people usually don't know exactly what's going to compel them what will solve their problem until the tools or raw materials are right in front of them. So here's the classic steve jobs quote on that point.

A

When I am evaluating tools, I like to start by reframing. What are we solving for good approach right for git ops? Let's do this through a frame of evolutionary neuroscience so that we can understand the cognitive processes at work at the high level. Software are tools that let us scale thinking this is what computers do for us. So we can ask the question understand what exactly is thinking natural selection has built up these cognitive and decision-making processes that empower individuals to deal flexibly and autonomously with new and unpredictable situations.

A

Another factoid about great apes is probably helpful here. We now know that all great apes understand a lot of their physical and social worlds in human-like ways, there's been a lot of research that demonstrates that they understand causal and intentional relationships between other creatures and things with the same exact brain processes that we use.

A

That means we can't say that our society, our culture, our language, has given us the ability to be able to think like humans. We have to flip that we share many of the important attributes of the way we think with our fellow great apes.

A

This also means I can use a chimp example to demonstrate the basics of thinking through an unpredictable situation.

A

A chimp might recognize that, because of some physical reason, a given situation demands that she use the only tool available to her and she needs to use it in a way that she has never tried before to achieve some goal sound familiar, so the chimp goes through. This process set the goal. I want to get that honey or I want to build this feature and before they act, they imagine what would happen if they tried different actions in the situation or if external forces enter the situation.

A

So, let's see say um swarm of bees come and get in your face right when you're about to get to the hive. So great apes essentially simulate in their minds what they might perceive when making the attempts to reach the goal.

A

Chimps do this: we do it too, but if I've learned one thing from engineering, things only gets me 10 of the way. This is my achilles heel. What I immediately loved about git and later about get ops is, if you can think it, you can try it and see it immediately being in this feedback. Loop gives me the ability to try things, while tracking all of my thoughts, what I'm trying to test and simulate along with the history of my work, so I don't have to go back and try those things all over.

A

It scales, your brain, so that you don't forget what you wanted to try what you tested, what you learned as a side side note when we were working on feminist, we realized that the best existing way to gather context from developers. It's definitely through commit messages.

A

It's it's a it's a good way. I mean it's the best current way, because it captures that context during your engineering workflow, while essentially, everything is being simulated in your brain and by extension, in the server so get ops provides this ideal environment to solve programming problems in that flexible individually, self-regulated style of cognition that also uses git tracks and controls our experiments so that we can use computing to make thoughtful programming decisions that serve our interests or that lead to working software.

A

This gives us the freedom to fake it until we make it and achieve our goal and in git ops, if we don't make it, the system can easily recover in real time. So it's almost like that experiment was imagined and never really happened. Ideally, michael tomasello, the preeminent pioneer of research into the origins of social cognition.

A

He calls this kind of process. Individual intentionality and I'm saying git is optimized, for this tomocello identifies three prerequisites that an organism needs to be able to act with individual intentionality. When we have an idea, we need the ability to cognitively represent experiences to ourselves. Offline.

A

Imagine and computing lets us do this as an extension of our brain. The next thing that we need is the ability to simulate or make inferences, transforming the representations of our ideas, causally intentionally and or logically gitops. Lets us do this in this case, with offline consequences to our decisions and the last thing we need to be able to self-monitor and evaluate how the simulated experiences might lead to specific behavioral outcomes where we really get. This is from ci cd get ops. We can test and understand the real outcome.

A

This is the brain science behind why developers love get ops. Achieving your individual intentions is where the dopamine comes from. This is why engineering leaders who want happy and productive engineers also get excited about this, because get ops helps us harness individual intentionality more effectively than ever to get that creative ingenuity, and this is what we're talking about when we say developer experience when I'm in an individual intentionality loop and the context is hot.

A

This is the ultimate yolo feeling it's so exhilarating to act on your calculated intentions and get that payoff now git does provide the sanity of a single source of truth better than any tool. I know, but we know that at least in our current state this does not hold up to our complex social institutions. Within our energy.

A

There are some cons to get ops. There are definitely workarounds. I've listed some of the cons here, but workarounds are expensive to implement, like scripting out a retry mechanism for the issues that inevitably arise when many automated systems are all trying to pull requests at scale right.

A

Policy automation is a very promising way of addressing some of these problems and making git ops more compelling for enterprises doing user research. For my policy product pointed something out to me. Policy helps us solve for something that michael tomasello calls joint intentionality.

A

What security engineers solutions, architects and fin-ops folks really wanted from themis was the ability to set policies for what mattered to them and ensure that they would govern the developers actions without requiring confrontation or conflict.

A

An important goal of the product was to provide gentle reminders within the workflows that would help developers see outside of their individual intentions, to deliver those features and into shared intentions around meeting the business needs. This really is a very human-like thing. Our software was trying to do when he talks about joint intentionality. Tomasello explains a lot of about inferences.

A

He explains how our cognition evolved around being able to understand implicitly why others are pointing out information to us why someone may or may not want to collaborate with us to accomplish a task, how to optimize our actions so that others will be better equipped to help us achieve a shared goal.

A

Lots of lots of complex inferences that we have to and a good policy driven product should do that, for you, no more overhead guessing what security wants you to do, reading between the lines on what can squeak by and what can't policy limits, how much we have to think about others intentions, because it bakes their intentions into the platform and ideally as engineers. We accept this because we know it's what's best for the enterprise, so that's policy, and I will say I found by working at devops companies and starting my own at policy.

A

Automation is another thing that gets very hard very quickly when you're, not opinionated, about tools. Applying policy to heterogeneous sdlc is not a solved problem and without policy get ops does not solve for joint intentionality. It's not optimized. For that, and I say this is a mistake.

A

An alternative title of this talk is really how it took me, two years of mentorship and work, to really learn that kelsey hightower's right and he's not the only one to zoom in on this brian dawson at cloudbees is calling 2021 the year of human centered tech. I've seen the symptoms of why this matters for a while. So when we optimize for our individual intentions without solving for joint goals, someone in the software supply chain suffers, someone ends up holding the bag for us and it's not pretty.

A

This is why I hear so often that the most serious problems in our field are the social and cultural ones, but this got me thinking in the year of human centered tech. What is it that actually makes human?

A

What is unique about human thought, and how does it relate to the work that we do, but speaking of the work that we do, I'm going to press pause for a second and I'm going to hop into a new cloud hosting platform to start up on my sql server for my demo.

A

Toolbar I'm going to go your dashboard and I am going to create a new private service.

A

Got this uh forked this from the friendly folks. uh The render docs.

A

Just checked this out this morning, all right, so I've got this repo that defines my sql server. I'm going to add.

B

Environment variable.

A

Oh yeah ah duh, I gotta name my.

A

A

It's simple here: I'm going to create the most secure implementation.

A

Possible and then I need to add a disc here for some storage cool stuff here around secrets, but yes, fairly.

B

A

B

Create the service.

A

All right so back to the presentation, just gonna, let that tick for a minute. So what we were talking about was what is unique about human thought.

A

So I did some research and it's surprising so there's a recent study that isolated the brain regions that are most active when we program this study was just published in 2021 out of nada japan, which is full of deer roaming, all around town. If you've ever been there, it's kind of fun all right, so we've got um three main sets of regions that are most active when we program we've got broca's area and for broca's area.

A

Findings suggests that the building blocks for that left hemisphere, dominance in speech, production that comes from that area was already evident at least five million years ago. So it's not unique, unique to human evolution.

A

This part of the brain is what we use for semantic knowledge retrieval in a goal-oriented way. So, for example, this is the part of the brain. That's extra active when you are listening to a sentence that has ambiguous words, homophones or a variety of possible, meaning you can see how that kind of differentiation would be really useful when you're programming.

A

Next, we have the left inferior, parietal lobe and the left supramarginal gyrus. This we use for episodic memory retrieval and guess what apes also remember in an integrated fashion. What? Where and when so how long ago, an event happened?

A

They have the same capacity and finally, we've got the right middle frontal, gyrus and the right inferior frontal gyrus. These deal with executive attention. Executive attention is the ability to control which sensory stimuli we attend to. We pay attention to and how we regulate our responses to those stimuli, especially in cases of conflict or fear when that limbic system is activated and the complex behavior of non-human primates, when given tests to assess these capacities indicates a lot of psychological continuity with human levels of cognitive control.

A

So all three of these brain areas and competencies are present in other great apes, present and being used, and experiments have demonstrated that these species do shockingly sophisticated cognition in these areas.

A

Now I'm not saying programmers are not highly skilled or that other apes can build software because obviously they can't, but my point is when we use our tools and our cognition to program we're not necessarily tapping into our fundamentally human powers, we're just not there's something more out there and you think a lot about what computers are good at versus what humans are good at.

A

I want to double click on what humans are good at, but before I do that, I'm going to step out for one more second to kick off some servers doing what they definitely do best deploying my app for the demo. So, let's see, if I'm ready to create a new web service- oh I'm still in progress, oh nope- maybe maybe I'm good.

A

So I need this string here. This is my server. I'm gonna refer to kinda really, but it's a service address. So now I'm going to go to create a new web service and I've got my repository. My github account some of the repos are linked here, so I can grab that right from the repo and call this quick gig, and so with render you can use um you. Can you don't have to have everything dockerized on your own, but I can talk about that in a.

B

A

I have a dockerfire file in my repo, so it's going to because I've selected this dock, but I'm going to add an environment variable here for my database uri.

B

A

I know pi my sequels. Let's put this.

B

B

A

My secure situation here.

B

A

Good all right, so I've added an environment variable here that will be present in my environment, as I boot. I don't think, there's anything else I want to customize here, although there's lots of options to do so, create a web service, and we will check in on that back to the presentation.

A

What we were talking about is what are our fundamentally human powers if they're, not those programming powers, well, ants can build an anthill right, but cooperation itself does not create complex cognitive skills.

A

Humans are unique cognitively, because the common ancestor two humans and other great apes had already evolved highly sophisticated skills of social, cognition and social manipulation which they used to compete, as well as highly sophisticated skills of physical cognition to manipulate causality in the context of using tools.

A

So our skills that help us understand how others other humans, particular goals and perceptions result in particular actions. Those were the starting elements from which we evolved even more sophisticated processes of joint intentionality involving joint goals, joint attention and those were built for focusing on coordinating with another social partner towards a common goal. Another example here, chimps forage together in small groups.

A

Let's say they come upon a fruit tree. Each individual chimp is going to scramble up on its own, find a good place to grab fruit on its own grab the fruit on its own and then even separate a little bit from the others to eat on its own, and when food conflicts arise, they turn into contests of dominance and that's it human foragers.

A

Conversely, as individuals um we only produce well in human forging societies, individuals produce the vast majority of their daily sustenance collaboratively with others as a rule and thomas has some theories on exactly when the human turn towards this collaboration happened. But at this point what we need to pay attention to is at some point in our evolution. Humans began a lifestyle in which we as individuals could not procure our daily sustenance alone. Still true right, we became interdependent with others to forage.

A

This meant that individuals needed to develop the skills and motivations to forage collaboratively to avoid starvation, and there was this direct selective pressure for skills and motivations for joint collaborative activity, also known as joint intentionality.

A

Also, humans began to evaluate each other as potential collaborative partners. Choosing a bad partner meant less food, so we became socially selected. We had to worry about evaluating others and how others were evaluating us. This makes a lot of our thought processes. Make more sense to me, but my point is joint intentionality is the fundamental human strength, social coordination based on overlapping intentions.

A

These are the problems that we were born to solve. Our strength is cooperating and being greater than the sum of our parts, and this is where we can shine far beyond a species: that's optimized for competition.

A

We have the power to solve together and manifest our joint intentions, but we have to focus on that and when I ask what does it look like to focus on that, based on what I also discovered about tool use and led to another piece of kelsey wisdom? Don't get lost in the tech? He did.

A

This talk at um portland devops group about spinnaker last year, and one thing that really stuck with me was this advice: focus on the front door, put the most thought into the way that your developers interact with the platform that they're using to ship, and he was right.

A

That's how you make the most of individual intentionality, but beyond that, making the experience uniform, making it painless being opinionated enough that you can run policy under the hood and no one even notices and keeping the integrity of the front door so that no one even notices when you switch tools, one notices when kubernetes is replaced by nergberg.

A

That's the gold. I think that's the gold, that's how software saves the world not currently happening right now, but it's possible.

A

All right, let's see where we're at with our with our demo, which you've already figured out, is a remade.

B

Remake of a great all right, so oh we're live. So let's see.

A

This is my service here and just to prove to you that my database actually works well. Logging in is not going to prove it to you well here my database doesn't work or my database works. Geez freudian all right, um I'm going to sign up. So I've got my name.

A

I'm going to dox myself here, see my email, no rosalind, rosalind.com, don't email me and my password is.

A

Register- and here we go- I have created an account, my application works, and all of that happened here in the background. While I was talking to you about what really matters to me, there's a lot of great features in here that I don't have time to show you today, so I want to go back here and wrap up.

A

My question for the cd foundation today is: can we build an open standard for this front door because that's the real goal right there? Let's help each other figure out how to scale the front door reliably. Let's do it together with the joint intention of saving the world through software, because that's possible.

A

Thank you for your time today. You can tweet at me about this at uh danilosaur. It's right here, that's a zero, not an o, and also I have a bibliography here, that'll be in with the slides and the book. I highly recommend is this michael tomasello's, a natural history of human thinking, thanks very much.