Brief Summary
The presentation discusses Canva's journey in evolving its infrastructure and platform engineering over the past four years. It emphasizes the shift from a reactive, order-taking approach to a proactive, product-focused mindset. Key aspects include building empowered teams, defining clear product visions, understanding user needs and stakeholder requirements, and balancing innovation with maintaining existing capabilities. The goal is to create platforms that not only serve the current business needs but also anticipate and unlock future capabilities, enabling product innovation.
- Empowering teams to own problem domains, not just solutions.
- Balancing qualitative and quantitative metrics to measure success.
- Investing strategically in new capabilities, improvements, and experimental frontiers.
The speaker, FAS, Head of Infrastructure at Canva, outlines the presentation's focus on the lessons learned in driving culture and training through platform engineering at Canva over the last four years. The aim is to summarize the journey, the experiments, and the resulting insights.
Traditional infrastructure is often perceived as static, inflexible, and slow to change, with issues only noticed when things fail. Modern cloud infrastructure is more flexible but introduces complexities due to distributed architectures. Many organizations are still evolving in their understanding and management of cloud environments. In 2020, Canva's infrastructure team was composed of skilled individuals who were primarily reactive and tactical, addressing immediate business needs as they arose.
A platform is defined as a foundation of self-service APIs, tools, services, knowledge, and support, arranged as a compelling internal product. The presentation transitions into discussing the specific steps Canva took to evolve its infrastructure and platform engineering.
Empowerment comes from teams owning the problem domain rather than just the solution or tool. Teams should have alignment on purpose and autonomy on how to deliver, underpinned by accountability for results. Empowered teams build great products that solve user problems while satisfying stakeholder requirements.
Users are those who depend on the outputs of the team's work, while stakeholders have influence and specific needs regarding the capabilities being built. For example, users want faster deployment times, while stakeholders want secure deployment processes with signed containers.
A clear vision is essential, but it should not be constrained by current tools and technologies. Internal product teams should define their own measures of success, distinct from the business's external metrics like revenue or user numbers. It's crucial to understand how users feel about the platform, not just rely on dashboard metrics. Qualitative feedback, like surveys and engagement with product managers, helps identify areas of friction.
Canva invests its platform engineering time in three key areas: 40% on building new capabilities, 40% on improving existing ones, and 20% on "KT," which seems to refer to experimental or exploratory projects. This distribution helps balance innovation with maintenance and improvement of current tools.
A platform strategy should avoid creating bottlenecks and instead enable teams to solve problems and innovate. Canva balances empowerment and enablement by providing "paved roads" (standardized solutions) for 80% of common needs, extension points for 15% of cases, and a "Wild Wild West" for the remaining 5%, where teams can experiment with minimal constraints but must own and manage their solutions. Capabilities developed in the "Wild Wild West" often get adopted into the standardized platform over time.
Canva's infrastructure has evolved into four platform organizations: Data, Cloud, Developer Platform, and Reliability. The goal is for the platforms to run ahead of the business, unlocking new capabilities that inform product development. This proactive approach allowed Canva to pivot quickly and capitalize on new technologies like generative AI.
Summarize any YouTube video in seconds
Get summaries like this one for any video, in your language, with the free Summ app.
Transcript
Full transcript of “Building Infrastructure as a Platform”. Click a timestamp to open the video at that moment. Need another video? Use the YouTube transcript generator.
0:00 [Music] hello everyone uh thanks Dave and and and everyone for for joining of course of course there's an update when I'm presenting there we go all right um so I'm FAS I'm head of infrastructure at canva my journey at canvas started around four years ago um with a very and I'm going to talk about lessons in driving culture training through platform engineering which is essentially in the next 25 30 minutes I'm going to try and summarize the last four years of the the journey we've been on what we've tried and and sort of some uh learnings out of that so let's start I love to start with definitions so we'll start with what is infrastructure
0:45 and if for folks outside this room the the person on the street uh they will give you that picture um things that are traditionally something like this static inflexible um rarely changing artifacts that take a long time to build they're very expensive and most people don't notice them until they don't work like the right Light Rail this morning which made me walk from Central instead of jumping on it um and while our modern Cloud infrastructure Stacks look a bit more like this and a lot more flexible the complexity of the architectures and the distributed model in which our software now works means that some of those difficulties are starting to uh creep back in and organizations in particular in how to hold the cloud um are still learning and still still evolving in in in their Journeys as as
1:31 we've been hearing about all all morning so this is or was infrastructure in 20120 uh when I started the the org was very nice and there was a few teams there's a few more in there but uh I picked the the four from memory um really great people knowing the the Technologies knowing how to build all the various things but very much being ordert takers and and you know reacting to the needs of the business being very tactical and like oh we need this type of database let's go figure out how um that particular uh piece of infrastructure isn't scaling let's go figure out how um and what I'm going to try and summarize next is to sort of go okay over the last four years how have we evolved that sort of very reactive technical nature and a little bit pivoted towards thinking about products
2:17 and platforms and and some of the uh things that we did in the organization to enable that all right another definition platforms we should sort of align on on what that is and there's many definition pling around I'm a fan of this one um by Martin Fowler and Evan butcher there um a platform is foundation of self-service API tool services knowledge and support which it arrang as a compelling internal product highlight on the product there um so yeah let's dive into it what did we do here's what we did the very first thing build Empower teams um and what do we mean by that the empowerment comes from the fact that the team owns the problem domain not the solution not the tool but the problem domain um we are on
3:05 GitHub we use them a lot but tomorrow might be something else the five years from now maybe something else entirely but the team that looks after Source control needs to be fully in charge of that problem domain not necessarily Solutions now with that I I thought I'd put in there but it's too dated that the Spotify um metrics if you remember that about alignment autonomy we talk a lot about that in our team have aligned on purpose and and what they're trying to achieve with the autonomy on how to go forth and and deliver on that we talk a lot about the third one in in at canva in our teams around accountability that underpins all of that that unless you are accountable for the results and and we're holding you to those results then a lot of the the the top two don't really don't really
3:51 matter so we've come with this definition Empower teams what did they do they build great products that solve user problems while set iying stakeholder requirements it is very important to know the difference between the two who are your stakeholders and who are your users um but before that or or as we break down that definition uh let's talk about product so it's something that solves a problem for users they're they're the key people here and something that provides value to them but last but not least it's something that also meets the stakeholders needs so again uses and St stakeholders users are the folks who depend on your outputs of your work and you're solving problems for them that they're utilizing things you build and your stakeholders
4:38 they hold a certain influence and and need some of the capabilities that we're going to build and and want to influence the characteristics and attributes of how we go about doing that as an example there so users they want to see a reduction in the time required to safely deployed production like how fast can I get my change from idea to to the customer's hands while stakeholders have a view on weight how do we do that securely well signed containers are the only things that are able to be deployed so we've talked about product we've talked about users we' talked about stakeholders but those empowered teams do they know where they're going and and how do they do that and for that you need a very clear Vision one thing we found as we're trying to build that vision is engineers in particular and
5:23 and very uh engineering focused teams love the current tools and Technologies they're using they only will have that frame of reference and that gets very hard to decouple from that and gold now let's talk about 3 years 5 years in the future what would amazing look like here um and that's what we tried and it took a long process but we but we got there in trying to build that CLA Clear Vision of we need to be informed by the current challenges and the current tooling but not be constrained by it and and look really further ahead while doing that we also need to be very clear on defining the SU successes of those teams now canva as folks would would know I I I'm assuming is a very product-led business and and has an amazing product which found
6:09 product Market fit at the right time and and is fantastic at that and yet our internal product teams their measures of success will not be the same measures that our product teams use so we're not going to measure in terminal products by by revenue or or users or or monthly actives or anything like that we need to own our own successes and what that means and Define that and be very clear about community communicating that to the rest of the business so you've got your metrics but I I like this question which I asked our teams If a metric changes as a dashboard is no one is there around to feel it or look at it does it really make a difference and the answer is really not really it doesn't because and I've set this challenge to a lot of the teams that I talked to a lot of leaders in my team that yes your your dashboard looks great but what are the users feeling
6:54 like we may have reduced the the time to deploy or reduced the CI CI time but but for authentication as an example our Engineers are having to touch the UB key like 30 times during that flow that is not a great experience so what we've learned is that you need both the qualitative metrics of yes numbers on a dashboard of of how long things take but also a qualitative feedback through surveys through um engagements with our product managers with the users to under really understand what are the challenges and where areas of friction still uh still exist so we're trying to do all that we have a vision we're building new things we're trying to improve capabilities but again can you change an engine M flight of an airplane not really not without having a
7:41 bunch of grumpy passengers because they don't like what you've done you've in you've introduced turbulence so what we talk a lot about is where are we investing our time that is a guardrail do not take that as gospel every organization is different but that's what we've found in terms of balancing where we invest our and we call that our platform P of where we put the work in so about 40% is building new capabilities about 40% is improving the existing ones making sure the things that toolings that people are using today are are being maintained and being improved and then KT is that uh 20% bucket which we which we always take off the top all right everything everywhere all at once then we've talked about all those things it's a great movie but a very terrible platform strategy because
8:27 what tends to happen if we follow all the stuff that we've talked about is you end up creating bottlenecks um and that is not what platforms is about platform is actually should be an Ena function with Empower teams solving and innovating for users reducing their cognitive load and by building great products with clear measures of success that is the the the future that we want to see where platforms is really enabling teams and I always tend to talk too fast so I'm realizing I'm near the end of my size but that's fine cuz we're in we're at lunch which is great um so this is the the the balance in how we've actually talked about and think about our empowerment and enablement which is that 80% will be the pave roads which is the stock standard way and we've talked
9:13 heard from Arman and others in the morning about templates and that's what we've built of if you want to start up a a new Java backend service that has a standard database and so on you can do that in a few clicks and you'll be Off to the Races um so make the complex things simple is a is a Cana value but I also like to add make the impossible probable and and hopefully we're doing that the 15% are your extension points of like oh I want to go off I have a standard uh Java backend service but I want to try out this new database sure we will'll allow you to do that will enable you to do the add-ons and last 5% are the New Frontiers or the Wild Wild West I like to call it we'll underpin and we will give you the guard rails that are secure so you can't shoot yourself in the foot but beyond that as an example provisioning a databas account with all the right things you
9:58 own it you build it you break it and you buy it right um it's completely out of warranty and and you will go do that now what we've found over the last four years is allowing people and empowering to do those things to the right side has created platform capabilities that eventually we adopt and bring into the left side into the 80% because they've figured out the rough edges there there's more uses across the business that that come up and and we pull those in so infrastructure in 2020 looked like this um today it looks a bit more like this um so we have four platform ORS underneath that the the various leads are there um data Cloud develop platform and reliability and we love to say at canva that we're only 1% of the way there and
10:44 we are only 1% of the way there um we are improving our platforms all the time improving the capabilities and my dream at least at canva that I like to say internally and I'll share with you folks is that I want our platforms to run ahead of the business we've gone from being order takers to servicing business as it is I would love for us to be in position and we're on the way there to unlock new capabilities that then informs a product to go hey did you know this was possible and we've seen some of those things um come to fruition with the the Gen and everything over the last couple of years coming out this allowed us to Pivot relatively quickly yes there was there's never no such thing as zero work involved there's always work involved but there was a much lower lift than we otherwise would have done had this all happened in early 2020 as an
11:31 example now if you've heard canva in the news and at the end of end of I won't end with it with a wrap so don't worry about that but I what will I what I will end with is a little bit of hype video on all the things that we've done over the last four years and how they've come to FR aition so thank you for listening I always end up talking too fast so I haven't taken up much every time but hopefully that leaves you free for lunch and with that over to this and down and [Music]
