Colin Humphreys, Syntasso. Anti-Patterns Exposed: Building Platforms the Right Way

July 10, 2023

Season 1, Episode 22

Excerpt here


0:16Welcome to cloud unplugged, thanks for being here on the podcast I know if you want to introduce yourself tell you a little bit about yourself awesome yeah thank you my name

0:23is Colin Humphries I’m the CEO of a company called centasso sutasso makes an open source framework called critics

0:29t-shirt here today some free advertising there yes it’s an open source framework called critics that we make it’s to help

0:35people build platform products so it’s a framework to help people build platform

0:41products I feel like every kind of word in that sentence needs to be decomposed unpacked we need to work out what that

0:46is but that’s what we do at syntase we build critics framework building platform products personally founded the

0:52company two years ago prior to that I was Chief technologist at VMware tanzu

0:57prior to that I was CTO at pivotal prior to that I was CEO of a little country

1:03called country company an entire country called Cloud Credo and the team at centasso where the leadership team we

1:09had at Cloud Credo and they’ve been through the pivotal VMware Journey as well so we’ve seen a huge amount

1:14platform space over the last kind of 20 years or so so we’ve seen and done a lot

1:20of the wrong ways to do things and we’re trying to now start doing things you know and Iota better keep learning from

1:25our mistakes keep adapting so we’re trying to just improve how we help people build great platforms okay so by

1:31platforms because it’s quite a I guess overused term in lots of different contexts so some people you know like I

1:38speak to somebody before and they might have like a they might classify their application as a platform because it’s got loads of components on it so by

1:44platform definition you kind of mean for developer like developer elements or

1:49enabling kind of developers to ship code and ship applications or I guess that’s the context yet this is a great question

1:55by the way because every term we use in our industry seems to be so heavily overloaded so platform’s a good one

2:01Services I think is the term that’s banned because I talk about services in the past and someone would say do you mean services so I’ve got my database

2:07and my cache or do you mean services like Professional Services like Services just means so many things to so many people like as a service platforms

2:14massively heavily overloaded so you do have to define the term but also even then you spoke about like four

2:19developers and I think it’s one of my kind of little bug Bears because like on the podcast I want to bring my my real

2:24self here and why don’t you see my real self I immediately start thinking this thing annoys me that thing annoys me so let’s just get this first thing out

2:29right now so I think often people use the term developer to mean somebody who works on a team typically building an

2:36application that drives business value and makes money and then they talk about somebody who works on a platform team

2:42almost synonymously with like operations and operators and my real passion here is arguably far more like platform

2:49development so I think an application team should have developers or people doing a development role people doing an

2:54operations role people doing product management role like design roles a whole bunch of like responsibilities on

3:00application teams and a platform team should also have developers on it so you’re saying it’s a platform for

3:05developers and I’m like yes it helps people doing development at the application Team level but at the

3:10platform Team level they need to have their own developers operators designers product manager measures they will rolls

3:18within a platform team so that we kind of rephrase them so the the consumers

3:23are the users of this platform Downstream user so they do like user research on what it is they’re building

3:29the user research should be performing for the platform team would be for the developers to find out what it is like

3:35how they wanted to work how they need to do things so the user dashing us would be of a developer for platform in this

3:41case I think there’s developers on the platform team and I think on the application teams there’s developers but

3:46also these people should be operating their applications are really good vernacular I think it’s been

3:52tremendously effective and I can’t speak highly enough about is team topologies but if you’ve read the team topology’s book but thoroughly recommend it people

3:58on the podcast I’m not I’m not going to Kickback but after saying this I’m going to try and get a kickback so teen topology is amazing book it provides a

4:04very simple vocabulary for how we describe teams and they talk about 14 types one of those team types is

4:10platform teams and another one which I’m saying application teams but effectively what’s synonymous with that is stream align teams that’s how they describe it

4:17and the stream online teams is the teams that drive the flow of the business and get the business value delivered so the

4:22role of the platform team is to make the stream aligned teams lives easier to be

4:28specific about that so they talk as well about this concept cognitive load how can we make life easier for the stream

4:34online teams how can we mean that they have to think about fewer things while they’re delivering business value so

4:40they can be more efficient high performance be more secure all of those kind of things so I think the thinking

4:45about in those terms is a very good way to look at it so they also give you the context by which to look at a platform team what that platform team delivers

4:52could be all manner of things it could be just Technologies could be tooling

4:57could be guides could be just so many things that they could deliver but their criteria for success is to what extent

5:05are they making life easier and lowering the cognitive load for those stream aligned teams I tend not to say stream

5:11aligned teams unless you’ve read the book is a weird concept I tend to say application team platform team but yeah so I had to unpick so much there to be

5:18like I’m talking about platforms it is this team who are or multiple teams who are helping out those application teams

5:24to make their lives easier yeah so I’m gonna be quite unpopular because I do love team topologies but I don’t like

5:29cognitive load yeah I don’t dislike it it’s like one of those things that does

5:34That’s So immeasurable it kind of is meaningless because I get it it’s like but I’ve never got to swim like your cognitive low looks really high right

5:41now and so it’s like it’s such a difficult thing to rationalize what you’re measuring against when you say it

5:46if you see what I mean whether it’s like productivity is a measurable thing if cognitive loads an impact of productivity then just say productivity

5:53I like that junk by the way I think I I a big fan of the whole if you can’t measure it you can’t move it and therefore cognitive load is tough to

5:59measure but I think when we lower it we can observe certain outcomes getting better so I often think a good way to

6:04talk through the platform teams I mean Dora metrics on the application teams are a good way of looking at it if you look at those how can we help them drive

6:11better durometrics to when their cognitive load is lowered they will be able to be more productive and then we

6:16will see these things get better but it feels like productivity is the thing because it could be that they’re just not highly skilled in that job and

6:23therefore the cognitive loads high and it’s nothing to do with the platform it’s just that they don’t have experience of the technology so it’s very difficult to be like oh well if we

6:29solve this but it’s like how do you differentiate I don’t know it’s got a charging thing I’ve kind of like it conceptually because it kind of makes

6:36sense obviously you know within lots of different contexts like having just a high cognitive load no matter what it was whatever you’re doing cooking it’s a

6:43complex recipe whatever it kind of is it kind of makes sense but it’s quite an immeasurable I don’t have an issue with

6:48it conceptually just an issue with it in terms of like business value recognition it’s like it’s can’t measure it I’m like

6:54I don’t know I just stick with productivity that’s how I feel a bit of a bit of a I could be properly massively

6:59unpopular people like because people I’ve heard people use it a lot a little bit of an outlier call it as you see it

7:04definitely like say this is this is just us having a chat on a podcast say like college you see it I do I mean what I can say maybe to empathize with you is

7:11that I definitely have an internal lookup table of Team topologies vocabulary to things that I would say to

7:17people in general unless I know they’ve read team topologies so I had to have this like look up so I when I generally tend to pull out like cognitive load in

7:23a chat with someone who I know isn’t familiar with the concepts I’ll talk about like their job is to Delight the

7:29platform team’s job is to like the application team to make their lives easier how can we help them like get to

7:34Value faster how can we help them have fewer failures I might start almost enumerating the Dora metrics or talking

7:39about being helpful and I think all those things are repercussions of lowering the cognitive load but so yeah I think just making life better for

7:45those teams that is a very easy way a very easy way to get to then you talk about how do we measure that and how do we move those needles but I also just

7:52want to come back to you saying about the word platform being so heavily overloaded the number of people we have where the we were working with we have a

7:58platform team but the stream online teams and organizations are also building a platform and it gets very

8:03complicated and also I’d say the majority of people I’ve been speaking to maybe the last five years have platform

8:09teams as I would identify them and don’t call them platform teams they typically call them SRE teams or devops team or

8:16support team or something else yet to all intents and purposes they’re what I describe as a platform team because their job is to make life easier for

8:23other teams in the organization so it’s been very interesting to see that and I think SRE teams I’ve never seen outside

8:29of Google an SRE team that SRE as per the Google book and has error budgets

8:35and the pages passing across so I think nearly all SRA teams are actually platform teams or even just operations teams and nobody I’ve seen outside of

8:42Google is doing SRE in the right way with the error budgets and the page of passing responsibility for that passing across based upon the downtime of the

8:49application its availability and doing engineering on the application itself because I think in Google they actually

8:55fix the problems to improve in the app itself being responsible like for the scale or reliability of it so they’ll be

9:01able to contribute whereas I think you’re right I think most of the time you just see embedded devops in an application team kind of Performing

9:08operational tasks but not necessarily being true by sres and yeah that does definitely see that so I think just to

9:13agree with you that I think our user vocabulary is it’s tough to communicate isn’t it it’s so tough I probably walk

9:19away thinking everyone’s understood me and probably I’ve always been just equally like I probably confused more people that I probably had sensible

9:24conversations but in my mind I said oh that went well and everybody Bamboozled by the time I’ve like walked away and

9:29probably like no idea what was going on about all these different terms I think we try to because after that I’ve Bamboozled with so many people I’ve been

9:35taught and tried to do a reflective listening if I repeat back to you and quite often those things become like cyclical where like you reflect back and

9:41they’re like that’s not quite what I meant but let’s try this out and you end up in like a circle of misunderstanding where everyone’s trying to reflect everyone else but it never quite closes

9:48the closest the situation so I guess then so the platform framework which is for them platform teams so I guess the

9:55team that’s made up of like a multi-disciplinary team is what you’re saying who wants to build a platform

10:01product is that what you would classify as or just the platform what was the classification at the end I suppose when

10:06you use the framework what does it give you so I think you could call it a platform product that’s interesting to describe it in that way I think calling

10:13it a I mean most people start with something they like to describe as a platform and then hopefully you’re

10:18evolving towards a better platform I think the product element gets super interesting to discuss so I’m going to

10:25make some controversial statements here put on a podcast why not so platform engineering is currently a very fashionable thing I’m a big fan of

10:31platform engineering but I see it absolutely as necessary building a good platform but not sufficient for building

10:38a good platform for all the reasons I just enumerated so in the same way that if I had a team and it was 100 engineers

10:45and I expected to build something that was awesome really delighted all my users I mean what’s going to happen next

10:50because they’re going to be lacking the skills around product management they’re going to be lacking design they can be

10:55lacking the components of what I call a balanced product team so whenever we talk about platform engineering it’s great let’s do some platform engineering

11:01because we’re going to need some Engineers to build a good platform but we’re also going to need product management to help us understand like

11:07the use cases Etc designed to help us have that empathy with the users and think about the experience we’re going

11:12to need to talk about all the parts of a balanced team so I think platform engineering is the first step towards

11:18platform as a product and having a product mindset so that we are building platforms that are there to address the

11:25needs of the users back to what I said in terms of like thinking about our stream online teams application teams as

11:30customers you know they’re the customers of this platform and we need to have that customer focus a dedication to

11:36delighting them all those parts that make platform as a product so I see platform engineering the current fad as

11:42just one sub-component of platform as a product and the other sub components are

11:47obviously product management and design and I think that the platform layer product management design are incredibly

11:54neglected disciplines and nearly every question I get asked about Platforms in terms of them not going well there’s you know things like the two-year Gantt

12:01chart building my platform I spent 10 billion quid on it I bought the platform and nobody used it and it’s like did you get fast feedback did you work in small

12:07batches did you do any kind of validation with your users I basically describe design and product management

12:12in terms of a series of questions to the people that have built these platforms and it’s just like no no no no no no no no no we sat down to two years

12:18engineering and no one’s using it in the business it’s a complete waste of money but we think we bought the world’s best platform it’s the most amazing thing so

12:24this is why I think platform engineering is great because you know you want to build actual value but in the absence of

12:30platform product management and platform design it’s going to be a wasted effort

12:35you need to do small batches fast feedback you need to learn with your users you need to empathize with your

12:40users you just focus on your users the two-year Gantt chart of spending a huge amount of money on loads of really you

12:46know building an amazingly fancy spaceship like those days are gone and good riddance to them and do you think then because we’re in the same space I

12:53suppose in some ways I’m not doing myself any favors but I kind of saying this but I can kind of understand like I

12:58suppose the investment cost if like from the business side if I was to wear take our hats off and then pretend I was like

13:05a customer and put that hat on you’re probably going to yield some cost benefits and cost value obviously from

13:11doing platform engineering but it is then a constant cost center because you can never resell it so like other

13:17products Investments the aim is it drives a revenue stream so having a multi-disciplinary team kind of makes

13:23sense because it’s going to be like potentially on most businesses it should be Revenue driven I imagine because it’s

13:28going to be kind of the business a platform isn’t directly Revenue driven it’s Revenue connected but it’s always

13:34then going to have a cost to it and an investment cost to it constantly so I

13:40can imagine people not you might not necessarily see it in such a way to think of it in the way that you’re

13:46saying because of that to kind of want to work in that way because you’re like why do we need so many people to

13:51facilitate the ask of like the rest of the business well I am going to make money for those products and why we’re treating this as if it is the same and

13:57it isn’t the same if you see what I mean thanks that’s a great question by the way I really appreciate that and I think talking about the business case I think

14:04that’s been one of the easiest things to talk about so my kind of lived experience with this is that way back

14:09when kind of late 90s when I started very much everything was devonops okay so you’ve got developers writing lots

14:16and lots of code and then they say they’re ready it works on my laptop they throw it over the wall to operations and

14:22then operations try and deploy it based upon some remember or something and the whole world falls apart okay I’ve taken

14:27Brands you I’m not going to name them big car brands big Airline Brands I’ve taken them offline for multiple days

14:32trying to deploy code and having it all break on me ejb1 from javrogb ones they went so wrong and huge huge outages

14:40across like really big Brands and everyone then points their finger at blame at everyone else it’s like when it works on my laptop or it doesn’t work in

14:45production who cares about your laptop I’m going to come and throw your laptop out of a window I’m going to throw you out of a window Etc that’s the kind of

14:50conversation you get into that didn’t work too well so then we had thankfully late noughties kind of 2008 2009 the

14:56move towards devops trying to culturally bring these groups together coach you think about driving shared outcomes together and that’s great so I think

15:03people really jumped into devops because it was far less toxic of a relationship between these kind of groups so what you then I think saw was you saw the

15:10emergence of devops teams in large organizations so I’d go to Banks and they’d have multiple hundreds of devops teams and what you’d see is effectively

15:18people who are in operations mixed in so cross-functional teams mixed into the development with the developers and each

15:23team would be assigned a small product and that was great and then effectively they had their own choice of what infrastructure you’re going to use what

15:29you’re going to use in terms of like you know your tin your everything like choose your Cloud choose whatever you want to do so I’ve seen this many many

15:34times multiple hundreds of devops teams but what you find is each devops team

15:40then builds its own little platform their own platform for their own application and when you saw like microservices come along as like let’s

15:46do that you saw each team may be building out like 20 microservices inside that one team not a great idea

15:51don’t want to go to microservices right now but you saw that happen quite a lot and then you saw they have to build their own platform to run those micro services and what you would see is

15:57multiple hundreds of very similar looking platforms emerging because you app can’t run without something to run

16:03it on so a platform needs to be there to run it so you see hundreds of platforms emerging and you typically see these

16:09teams particularly in large organizations where there’s lots of regulatory stuff compliance governance Etc all of that kind of stuff you see

16:15this team spending most 20 of their time actually delivering business value and 80 of the time working out how to get

16:21the platform and the infrastructure to be compliant to be governed to work properly all that kind of stuff so

16:26effectively you might be looking at an organization with say 5 000 developers and effectively 4 000 developers worth

16:32of time is on building platforms so you might take a hundred of those developers and build a great platform and then

16:38you’re going to save yourself like 3 900 developers worth of time now you don’t have to fire those developers you can

16:43put them actually onto building great business values another way to look at it is you’re going to 5x the

16:49capabilities in your business and when they’re delivering those capabilities because they’ve got a platform which hopefully delivers things as a service

16:55when they’re needed that are relevant and bespoke to the organization sorry listeners if you’re not watching you’re

17:00listening I was slacking my hand there next to the microphone I think what the kids do when they clap as a service

17:06bespoke for emphasis so when you get that platform that really is delivering great outcomes inside the business that

17:12means before when everyone would have a long time to Value because you’d have to like you get your team going you build your platform Etc now people are like

17:18yes I need some web serving I need database I need cash I need the identity I need this and they’re just making API

17:24calls and getting those things and just getting on with the business they need to be getting on with rather than thinking about platform building so when

17:30we see platform to be successful it’s tremendous business outcomes and I think you can talk about productivity going up

17:36exponentially you can talk about efficiency so that business if it’s kind of a steady state business I wouldn’t

17:42recommend this but they could fire the other 3900 developers you know have a massive cost-cutting efficiency has gone

17:48up but then you talk about security and risk as well because if that team of 100 are doing an amazing job building a

17:54great platform that encodes all the right governance and compliance and everything else into what’s been offered

17:59as a service from the platform that means there’s so much less risk in the business because people are using effectively standardized well-governed

18:06components to build their stuff you know the platform provides all of that and therefore your chances of a data leak

18:11your chances of you know being hacked all that kind of stuff is massively reduced in the business this so I think

18:17there’s so many ways by which platforms can bring tremendous gains to the business on so many aspects that I found

18:24that part of it to be you know very easy to talk to people about yeah because I’ve kind of said this before probably

18:29on there but because if you’re spot on with all those things I guess if everybody’s Reinventing the wheel there’s a high cost to that wheel

18:36reinvention anyway if you’re all solving pretty much the same problem and all the things you’ve just said all like the non-functional requirement elements

18:42around it or compliance requirements Etc but what is interesting as well on that is like the market when you look at the

18:48markets the cloud security Market is exponentially growing and that means there’s a bunch of symptoms being caused

18:54by something so the root cause isn’t like well you know I signed up for cloud we haven’t used it yet but we’re really

19:01worried about security you’re like well clearly you’re not you’re only worried about it after you’ve consumed it so

19:06it’s like so kind of what’s going on then on the consumption of it where you’re now worried about how you’ve

19:11consumed it so surely there’s a problem manifesting on like how how you are actually consuming the cloud for it to

19:18be such a concern that you now have to go and get something to tell you like whether you’ve done a good job or not you know and how you’ve configured it

19:24and what the risk is so it’s the preventative side for that cause is a thing that’s really that’s how you’ve

19:29done it is the problem really and that’s what platform engineering is to solve really is now I think as well yeah so

19:35much like following through that history I was talking about like devanops devops I think those devops teams and the

19:40combination of that with the forces are coming in from microservices they’ve caused a ton of problems well I mean

19:45they also did it with a lot of value if you just imagine I think what’s happened is those teams were focused on getting

19:50lots of code out quick and we’ve been able to throw out so much code very quickly particularly with containers we’ve just got like stuff running

19:57everywhere and someone came up to me a conference a while back and said this is when um heart bleeds remember like loads

20:02of SSL vulnerabilities coming out with all with amazing brands for coming out all at the same like one after the next quite a few years ago and the characters

20:08I’ve got 30 000 containers running and I scan them and vast majority of them have like SSO exploits all over them what can

20:14I do about this and the time I was talking about Cloud foundries I have a lot of experience with Cloud Foundry and

20:20kind of like pass as we knew it so kind of 12 Factor pads is like a rocker and Cloud Foundry and they said you know

20:25you’re talking about build packs here and you mentioned about if we had this kind of thing so if we had a customer way back in the day that had 200 000

20:32applications running all of which had an Estelle vulnerability and we swapped out a component called a build pack instead

20:37of applications and upgraded all of them just as is so because they were using a platform and one of the components of

20:44the platform was the dependencies such as SSL you just swap out the build pack and It upgraded the whole lot so that’s

20:50200 000 done in a single command now this person said I’ve got 30 000 containers they’re all of you know

20:56unknown provenance like they’ve just come into the platform what do I do about them and I said you can only contact the container authors set the

21:02sorts out so I think that emergence of you absolutely spot on with your point about the security tooling where

21:08security tooling is like taking off because we’re kind of causing ourselves a huge number of problems by not having

21:14great platforms that kind of take care of that for us you know there’s so many teams out there this is the cognitive low Point again like if you’re building

21:20your own platform you have so much to worry about particularly in a large organization there’s teams that have to take care of it in the cells how do they

21:25get anything done because you just got so many concerns that aren’t the business value you’re trying to achieve so that’s why I have a big advocate for

21:32people building great platforms to help those poor people that have to have to deliver the business value like get on

21:37with their jobs I do want to mention I sound like I’m just here evangelizing platforms platforms no I think it’s good though because it’s a market isn’t it in

21:43itself now I think it’s kind of become a recognizing the same as you’ve got the cloud security market and you’ve got like monitoring and logging and all

21:49these different things and it’s been around a while it’s not like it’s new I think it’s just because maybe the complexity of all the different things

21:56out there like there’s so much choice in which case like how are you integrating with all of this choice and who’s

22:01integrating all this Choice together into something that can be meaningful to somebody else so I think platform

22:06engineering as a thing is obviously just kind of necessary to function my own business but there’s the misconception

22:12of what isn’t an actual platform like you were saying and kind of product teams in doing that I think is probably

22:17the linchpin of like outside of measuring velocity and Dora metrics which are quite I think more

22:23productivity and reliability focused they’re not cost Centric so you don’t have like operational cost in that it

22:30doesn’t talk about scale of a team like what’s the balance you know if the Dora matches are amazing but I spent 200

22:36million is that good you know it’s like am I doing well and it’s like what’s the cost ratio of the Dora in context and

22:43security and other things because it’s so vast it focuses on an element of which is really important and then there is a whole other bit that’s kind of not

22:50yet recognized as much in the industry so I think some of the things you’re saying around the team and the makeup of

22:56the team and knowing your users and researching what they need and making sure and investing you know from an

23:01engineering perspective and making it right is really important because that’s the frame what it really is You’re

23:07Building at that point I mean and the importance of what it is you’re building because it’s got complexity to it and it

23:12needs to be thought about properly and needs to have the research in it to do it just this erasing 70 good points here

23:17John as it’s kind of like moving on to the second ago and I think there’s some bad things I’ve seen happen with platforms that may be a bad platform

23:24could be masked in amongst all the metrics we’ve just described and I think the key sign I think in a bad platform which wouldn’t get picked up by other

23:29metrics we just mentioned is when those application teams are forced to use it

23:35you see this happen in so many businesses that we’re going to build a most amazing platform we’re going to spend 20 million on it it’s going to be

23:40great they build the platform and they say oh spent 20 million on this thing now everyone has to use it Thou shalt use the platform this is bad for a few

23:46reasons so firstly it removes the feedback mechanism from the platform because you just got 100 adoption so now

23:51you have no feedback mechanism it removes any kind of learning because you want people throwing application teams to be trying out new techniques apis

23:58tools in the industry and saying hang on this thing’s really cool can we do a prototype with you on this thing and

24:04then we’ll start to use it on our team and then you can see if other teams want it and then you can use us as a sensor making mechanism and then take that out

24:10to other teams that’s how you start to grow your product API the moment you say to people you can only use the things

24:15I’m on the platform you can’t use anything else your platform is almost fundamentally going to be terrible from that point onwards it’s just it’s like

24:21bad platform start from mandates of Thou shalt user platform controversy would you say that the same as cloud like you

24:27shall use AWS would you yes I would the streamlined teams Drive the business in terms of the actual money coming into

24:33the business and they’re driving the outcomes for the business so they need to be the king makers they need to look we makers they need to be deciding what

24:40is of use to them if you tell them this shall be of use to you I mean sure they’ll use it it might be tremendously reliable but you might find there’s

24:46something else out there that can make them 10x more productive something else out there can make them much more efficient and they’re not using it

24:51purely because of this mandate so your platform’s ability to learn and adapt as we said earlier about being a good

24:57product the moment you mandate your platform shall be used it’s no longer a product it’s just a platform you’re just

25:02back to the world of platform engineering again this is why I think you remove your ability to learn and to

25:08create a continuously learning platform product by mandating people use it I

25:13think some of the more effective ones that I’ve seen you mentioned about how it’s not an internal revenue Center why not even if you create some kind of a

25:19token system just like some kind of chargeback something where you can see people are consuming these things I can see how much they’re consuming it so I

25:26can see if they’re using other things because they’d rather use those other things so you can see so some kind of notion of having feedback baked into the

25:31system I think you need that to build a good platform whether it’s real money being exchanged or not I think you need to have that sense that if we’re going

25:39to provide postgres as a service internally people have the option to go and use like RDS or you know Google

25:46Cloud SQL or whatever they can use whatever they want to use but they choose not to because the one we use internally is kind of wired into all the

25:52right places the data is going to be the right place for that gdpr and governance and all that kind of stuff so they know if they use our internal one they get a

25:59better all-wound package than something they take off the shelf that’s where the internal platforms can really add value

26:05because you know it’s the right thing for you to use and you’re choosing to use it because it meets all of those things that you need yeah that does make

26:11perfect sense it’s tough though to be fair though I mean not mandating I think it’s fine and you shouldn’t have a mandate because you just use it or like

26:18basically that’s it you only get one choice and the choice is my platform and the end but I can also see too much

26:25choice you know facilitating so much choice is also problematic in itself because as much as you want to be able

26:31to support you know the right thing across the business if there’s an element of scale on that then the choice is exponentially grown too there’s going

26:39to be an engineering cost to make sure that all of the things that you’ve got to do have got to be kind of catered For so you’re like that’s the trade-off

26:45again is like it’s hard because am I trading off now on do I scale up my team to support all the choice internally

26:52it’s just a difficult Balancing Act of like can I get more money please because there’s more Choice needed over here and

26:57we need to do a bit more engineering and it’s like okay yes but couldn’t they use something else because isn’t that just

27:02quite similar to this other thing we’re kind of using like well it is kind of similar yeah but it’s apparently better and thank you it’s a really good point I

27:08think we’re describing there again is product management product management is not saying yes to every feature request it’s near you’re always saying no to

27:14things and explaining why is a better way of doing these things so this is my point about like I’m talking about what goes wrong with platforms they often

27:19describe product management back to me and I thought that’s what’s Happening Here describing like good product management back to me I think if these stream online teams if they want to go

27:26like actually I’m not going to use these things I’m going to go and use these other things I think this comes back to another thing we said earlier but those

27:31stream online teams should not be just developers they should bear the operational cost of the decisions they

27:37made when they go off platform so I’d like to talk about this in terms of skiing which is a slightly weird analogy but like the ski runs are like your

27:44platform you’ve gone out there you’ve looked at where it’s safe you’ve put the lifts right next to them you’ve pee

27:50smashed them so they’re lovely and smooth and people can go and have a great amazing time on those lifts fantastic this is your platform it’s

27:55like your kind of pave path analogy it’s like if you stay on the platform you’re all good we know you aren’t going to get into trouble okay if you go off peace

28:02there’s Bears there’s rocks there’s trees how good a skier are you really and people need to feel that because I

28:09think quite too often they go and make some bad choices and they throw it over to operations and say yeah I’ve just selected this random bunch of stuff to

28:15use in my app and I just write the code and look it works on my laptop now you sort the rest of it out so those application teams they need to like feel

28:22the pain that they’ve incurred so they need to be not just application teams in terms of development they all seem to be application operators so if they choose

28:28to go off peace that’s their choice and there will be times when they choose to and it’s quite right to do so I work with a team that needed like 40 gig

28:35internal networking for their data couldn’t get out of the cloud had to buy some tin and they bore the cost of those decisions and that’s what made their app

28:41work so there are times when it’s the right thing to do because you need that functionality but they need to be aware

28:47that they just write this like bizarre over-the-top like rider for like the platform people where it’s like yeah I want M M’s I’ve never taken out you go

28:54and bear the cost of this they need to be aware when they take those choices about building esoteric things this isn’t about making their CV better like

29:01they need to that team and them they are going to bear the cost to those choices that they know yeah that’s really good actually because I think that’ll at

29:07least coerce behaviors on wanting to play around with different things but I guess I guess we’ve covered quite a lot

29:12of different various things because the requirement hand modeling which is like the product management aspect to

29:18probably do it in the way we’ve just kind of discussed on like well that sounds similar to this and this and so that I need the level of domain

29:24knowledge having domain knowledge and product management skills very difficult in our space I would say because that’s

29:30quite a hard it’s not impossible obviously I think people can learn product management skills but to understand engineering to understand the

29:37personas to understand the overall outcome you’re striving for and then the ecosystem that those developers are kind

29:44of surrounding like that’s a plethora of skills I think it’s too much to expect any one person to have those skills I

29:49think that’s why you need a balanced team very much the platform team should be regarded like the similar way you regard an application team with that set

29:55of skills I think exists in their one person and if they did they’d be such a generalist they wouldn’t be good at any of them therefore you need to like focus

30:00on constructing a balance team but I think your point your earlier point about not Revenue generating or not perceived as Revenue generating I think

30:06it’s some of the short-sightedness we see in business so they don’t allocate those key skills onto those teams it’s just the people that used to the

30:12operations are now relabel platform Engineers expected to get on with it and I think that underserves the people on

30:17that team who are going to need support in terms of those new disciplines also understands the business I mean I gave you an example about a business like

30:235xing their productivity and their efficiency who doesn’t want to do that and yet I think many businesses I speak

30:28to are too short-sighted to realize this but people there’s probably a misperception though I think because it’s so ill-defined anyway people are

30:34measuring things what you’re measuring against isn’t hugely expansive and I

30:40think most people and maybe people can definitely disagree with here is that

30:45tools people CI jobs is a platform right so I trigger a CI job it runs some

30:51tooling it’s got configuration to it you know we’ve got a team of people over it we’re a platform team because it’s kind

30:57of what I would call like manual automation where it’s got human beings in it it’s not really and that those

31:03human beings are almost like a gateway to this platform and not the actual platform isn’t really a consumable thing

31:10thank you it’s a really good point and I take the opportunity to plug something that’s on the syntastic website we have a little micro site called four Journeys

31:17and it talks about some of the like anti-patterns of platform building trying to help people understand where they are now and how they can move

31:23towards better so I think what you just discussed is one of those kind of anti-patterns so the key ones I wanted to mention I think a lot of platform

31:29teams fail by providing effectively the keys to a public cloud and then saying

31:35off you go here’s AWS access that’s it off you go or giving people like raw kubernetes and saying like you’ve got

31:40kubernetes just what kubernetes the case API just put whatever you want kubernetes we’ll make sure like the kublets are running and like the API

31:46servers are running in the scheduler and you just do whatever you want now so it’s entirely up to you so I think a lot

31:52of people hand over infrastructure and say like it’s done and what actually happens in those situations is because infrastructure is in no way customized

31:58or bespoke to the business that it’s in they’ve still got all of that cognitive load and all of those concerns are building their own platform they’re just

32:04building on something slightly different to VMS or in some cases just on VMS on the public Cloud so you’ve in no way you’re not a platform team if you hand

32:10out like raw apis straight over to the teams from public cloud or kubernetes you’re an infrastructure team let’s just

32:16call it what it is you’re the team handing out vsphere like that’s it you’re just an infrastructure team so then I think the reaction to that is

32:22people try and do higher level abstractions that are of greater value and then you end up people I would say creating things like big monolithic

32:28terraform code bases where it’s like who owns the code base who’s going to pull request in like we’re going to share the code base between teams isn’t that a

32:35good idea nothing bad could happen with that you see people like oh it’s fine you write that’s a jira ticket and then

32:40it’s fine we’re really quick with tickets and immediately six-week ticket queue happens so we see jira we see terraform I know what we’ll do to make

32:46life easier for everyone we’ll give you Kate’s access we’ll give you some Helm charts we found on the internet and you kind of use them when you want but then

32:53when you deploy that you need to maintain it it’s like this anti-pattern after anti-pattern of how people do things and I think I mean again team

32:59topologies like had this interaction models around kind of like you know collaboration but the key when a platform team needs to move towards is X

33:06is a service where you are going into your organization you are learning via collaboration what the right services

33:11are so you don’t start with a Gantt chart and a big plan I’ve just thought of of all the right Services you need to

33:16go and collaborate with application teams learn what causes them pain what the problems are in their lives and what services can be delivered and then build

33:23out those services with them in collaboration and then offer those things as a service within your organization buy as a service I mean on

33:30demand I think right now platform teams are stuck between either you give people things that are on demand like public

33:37cloud like the Kates API or you give people that things that are bespoke and relevant in your organization like the

33:44helm charts you’ve just customized like your terraform repo like you know you find a Geo ticket and I’ll go and do a bunch of like SSH commands and run books

33:50so it’s like either on demand or bespoke what organizations need to move towards building a good platform is about

33:56offering on-demand bespoke services that are high level for their organization and I’m sure someone’s making a

34:03framework somewhere to help with that I can’t quite remember what it’s called is

34:09it krantix critics something like that and that’s it’s like key point so you

34:14know we’re kind of again we’re only a small startup we’ve been founded based upon what we’ve learned on these Journeys with people and we just don’t

34:19see anything that’s out there that is about helping people create bespoke but

34:25on-demand services in their organization and I think that’s really what like the thing that drives me is why I wake up in

34:30the morning is try and help companies build out that really like high level productive API of services in their org

34:37that you know aren’t a jira ticket aren’t a shared terraform repo on a bunch of Helm charts here’s the keys to

34:43AWS and off you go like those things I think are all the wrong abstraction and one way of thinking about it used to be

34:48high level bespoke and proper on-demand API yeah that’s really good do you think

34:54then because that sounds like I kind of understand exactly now what critics is to be fair now kind of cemented into me

35:00now you’ve kind of contextualized it the way that you have and does make a huge amount of sense to me the flip side I

35:06guess outside of like do you think the reason those behaviors or the ways of working or existing like the

35:11anti-patterns that you’ve kind of Labor or do you think that is because there there isn’t the right things in the

35:16industry to solve those problems or do you think that the industry itself is

35:22kind of slightly Echo chambered around you know what because I guess like any like any industry in any Market there’s

35:28a huge amount of traction with certain tools like the things you labeled you know around like hell and you label like

35:35terraform and you know kubernetes and so there’s a bunch of technology that

35:41people just hear and they’ll hear it all the time and therefore you’ll assume and they could these could be Technologies

35:46even in you know people talking around like kubecon and cncf or platform con and all the different flavors of all

35:52different things so you naturally might assimilate that that is platform engineering because like yeah I’m doing

35:58like my peers use these tools and they’re doing platform engineering and I’m using those tools so I’m doing

36:03platform engineer we’re all building platforms and yeah that’s it that’s what platform engineering is and I guess it’s

36:09a harder thing because you’re a bit probably similar to in thought to me you’re contextualizing business value around it be beyond the technology kind

36:16of influence right because the Technology’s there but they’re tools more than platforms and then in your

36:22community from a business perspective and being like well something needs to coerce things into something for the

36:27users so they can actually accelerate what they’re trying to do it’s kind of meaningful to them but our industry isn’t necessarily that market isn’t

36:35necessarily designed for that thinking at the moment hugely I would say maybe he’s going to get there I don’t know if

36:40you agree with me over time and that’s kind of you know the platform engineering emergency and what that’s all about but I don’t know if it does or

36:46doesn’t really help convey that message very well thank you it’s really interesting I think it’s tough to reflect on this and try and work out if

36:52what we’re doing is useful why hasn’t it been done before in quite the way I mean we can only each talk about our own

36:58Journeys and our own experiences and then share those Journeys with other people and share those experience with other people and see how they kind of

37:03land because you mentioned how like what I was talking about it’s kind of landed where were you and made sense I do want to say that of the tools we enumerated I

37:10like all of those tools by the way it’s more about like what we Define as being a good platform so it enjoys those tools is definitely a place for those tools

37:16they’re all better than what came before just to be clear my personal journey I mean way back when I did everything by

37:21hand and then I built sort of 2004 I bought out Volkswagen’s infrastructure using CF engine so CF engine version one

37:27that was my first kind of infrastructure as code bike set of things to build out all of info swag and stuff and then I

37:33kind of followed the journey so I did all of eBay Europe using puppet and that was like again a Learning Journey in evolution much much easier to use than

37:39CF engine CF engine early days it was like almost like a theory rather than a practice and then built up things like

37:45cineworld Cinemas using Chef so I really like Chef again as an evolution after puppet so kind of go on this journey

37:50together and that was kind of taking things from the more kind of infrastructure as code approach then Cloud Foundry kind of came out and

37:56really generalized a lot of those things so it gave you the ability to take my code as long as my code obeys kind of

38:02the 12 Factor Manifesto obeys a certain set of rules I can just push my code into Cloud Foundry and it just runs you

38:08know Cloud Foundry borrowed a lot from Google a huge amount underlined technology was like I mean they just

38:14seems kubernetes these are just all the ideas from the board it was built by team here x Google and it works super

38:19super well for that code so long as your code obeyed its rules and the challenge we then found was that you take that you

38:27have that opinion back to all of that you build the best possible platform with Cloud Foundry and you take it out to Big organizations and each big

38:33organization wants to do everything slightly different they all have a slightly different set of regulatory concerns and compliance and governance

38:39and the way their teams work and opinions about where data sits and how it all that stuff is slightly different so yeah everyone wants to do 80 the same

38:46and 20 different or less but the 20 they want to do differently is in a different place in the stack for each of them so

38:53you can’t ever have one off-the-shelf thing that meets everyone’s needs so this is my problem with like you know

38:58giving people Cloud accounts if every business use the same set of stuff in

39:04the same way AWS will be able to give you like the one true platform that works but they can’t they give you this massive wealth of services many of which

39:11are just infrastructure level things because every business wants to do something a little bit different with them so I firmly believe the reason why

39:17platform engineering and platform as a product more specifically are going to exist for a very long time is that your

39:23infrastructure is commodity your platform is a product and then what you offer your value that needs to be

39:29differentiated in novel so you can make profit on it from your organization so it’s like the stack so your apps need to be novel your platform needs to be a

39:36product within your organization and your infrastructure needs to be commodity that’s why you can take your infrastructure off the shelf you need a

39:42team to build your platform product and then that is what we celebrate your business so that your novel differentiated business value really

39:49drives value like as in people pay lots of money for it so I think that stack exists almost independent of the

39:54vertical that you’re in and that’s why I think people have really failed I mean you know CTO at pivotal we tried to

39:59build the one platform like pivotal Cloud Foundry that was going to take over the world and be everybody’s platform why didn’t that succeed why am

40:06I still not there now why am I not one of the richest people on you know in the world because you can’t build one platform that meets everyone’s needs the

40:13only person who can build your platform form is you so that’s why we’re now building out critics is because we’re

40:19trying to make it easy for you to build your platform rather than a vendor building your platform for you does that

40:24make sense yeah it does make a lot of sense I think with the 80 20 rule because I kind of listening in and then I haven’t just kind of reflecting back

40:31on things you were saying I’m like Google like Borg and the standardization but the value it bought Google was

40:37probably astronomical and so the problem with the organization are facing as much as I guess if I was being pragmatic or

40:45objective around the problem is that standardization will bring you more value even than the product platform as

40:50a product because why are they so different really like what’s behaviorally going on where people are

40:56making such different decisions like technology is diverse you know but

41:01actually when you do really look at what people are doing the logic the business logic you can standardize on that

41:07business the actual outcome that what they built gave to the customer in the end is not so differential each time

41:13where every business of every single product of everything that is so unique we might have to agree to disagree on

41:18that one because that was my lift experience at pivotal speak I think the nature of that deviation in the scale of it is maybe proportional to the size of

41:25the company pivotally targeted like the largest companies in the world like top thousand largest companies and each one wanted to do things kind of like do

41:31things differently so you had a magic one though that wasn’t about I mean just taking away platforms and products and all that at the equation and you can

41:38look at it more on the problem itself you know like you know does homogeny solve the problem I guess like if you

41:45were to make some concrete decisions about business around how things were engineered and you could standardize on it would that reduce the need for a

41:51platform at all but it’s like a hypothesis almost nothing to do it’s really interesting to kind of think about that and think why can’t we build

41:58a platform that everyone can use because you’re going lowest common denominator at that point in time because everyone

42:03has to be able to use it and that’s why your lowest common denominator is the infrastructure when you try and take that info level to The Next Step it’s

42:09almost like the differentiation sorry they took a step back here all businesses would make no profit if every

42:14everything they offered was homogeneous commodity there’s almost that definition of like so go to economics here like

42:19they’d make no profit if everything was just being sold at the kind of like marginal cost equals marginal price of

42:25Economics so you’d have like just have undifferentiated commodity for absolutely everything so the Way businesses make money is by

42:30differentiation and doing something a little bit unique now I think at the platform layer that differentiation starts to come in I appreciate what

42:37you’re saying in terms of like but the Technologies might be the same but everyone’s Technologies might be a little bit different or their business process more to the point might be a

42:43little bit different or their governance their compliance the way they set out might be a little bit different so people start to compete in these aspects

42:49where they do things a little bit differently and everything makes it that’s my point they’re mostly the same if we blow your eyes it looks exactly

42:54the same and when you go on the ground and you do it with these organizations it ain’t the same and you end up with everyone trying to decompile your

43:01software recompile in a different way compose it in a different way like everybody did everything just a little bit differently so those platforms are

43:08bespoke are unique and I say I appreciate you may have lived different experiences than I have but like my experience was I just found Intrigue

43:15because I totally agree because my experience is exactly the same so it’s like I guess what I’m doing is like almost picking at the problem like

43:20hypothesizing on like taking away I guess this the where we’ve come up from the industry because I think we fully

43:25align on like the need for the platform engineering and products and actually things are unique and things are different and businesses are organized

43:32that way I was just kind of like parking that and then looking at the problem because like just because it’s intriguing because it’s quite behavioral

43:37and People based and it makes you think like well what happens if AI is writing the code is it that different what

43:42happens when all this stuff happens like do you have unique like edge cases is the human element the bit that’s really

43:48causing the Divergence or is it really truly is actually the businesses requiring Divergence and I guess that’s

43:55really the tail end of it is like because it’s an interesting password on anyway where things are expanding in

44:03different ways obviously from Ai and you know obvious code and everything else and so co-pilot and things so you’re

44:08asking a really good question there and I think there’s sorry if I kind of immediately clamped down YouTube no it’s good I just like to explain where I was

44:14coming from because I think I was kind of often attended I knew what I meant in my head but before it was I wasn’t communicating very well but let me see

44:19if I can reflect this question in a slightly different way that really landed very well with me was I do think

44:25so many businesses invent things that aren’t actually different about them at all they shouldn’t do anyway they’re a

44:30complete waste of time so in terms of like should they be doing these differentiated values of the 20 that’s

44:35different maybe two percent is real difference and like 18 is just them wasting their own time but because

44:41they’ve misinterpreted some kind of like governance or regulation thing so I see that so so commonly and then the next

44:47part on top of that is how much of a performance gain would get if they didn’t deviate in that way so it’s

44:52almost like they may be deviating these ways that are like you know and the infrastructure is zero percent on the platform 30 and on the app 70 but should

45:01it actually be like zero percent a platform should you just use like Heroku off the shelf or whatever but I’ve seen

45:06so many organizations where we would deliver platforms and then people would have a massive incentive to use the

45:12platform because they could offload all these kind of like operational concerns onto the platform and yet couldn’t for

45:18all these reasons so it sounds like you invert it where rather than saying people are you know I’m on the platform team therefore I want to build a

45:25platform I don’t want something off the shelf if you invert it to the application teams have the platform there so use it if you use it it’s free

45:30you don’t have to worry about any of the Taps the app’s Staying Alive on that’s taken care of by the platform so let’s go offline once they’ve passed the test

45:36Suite we’ll keep them running for you just don’t you worry about it it’s so good for you and then you get some adoption but not total adoption and you

45:42teach people why not and they say there’s these things we just can’t get rid of them so I think there was a lot of like not invented here going on

45:47there’s a lot of like I want to build it myself going a lot of CV based development going on all those things are absolutely true a lot of

45:53misunderstanding of rules and regulations a lot of cargo coating or the person who came in before me used to do it this way therefore I do it that

45:59way all that stuff I absolutely agree I think that’s what you’re kind of your experience maybe you’ve been met those kind of people doing those kind of

46:05things where they have maybe invented like traps for themselves to fall into based upon their experiences yeah I

46:10guess I’ve just seen where the problem’s moving to because I think I’m totally in alignment and I agree I everything you’ve said is absolutely exactly the

46:16same experiences too and I’m also not saying that when it doesn’t fit if you

46:21can’t modernize it or change it then if therefore just basically just should stay there and you can’t use the platform but obviously that’s not an

46:27answer either but it’s almost like the problems moving around I mean it’s like there’s a root causes of figuring I

46:33think what you just said is like was it really just two percent rather than like the twenty percent and we can’t solve that with a platform at all right and

46:40that’s behavioral and it’s organizational and it’s very difficult to solve and these things happen for a reason and there’s decisions made no one

46:47can rewind time and get into a Time catchable and be like well we need a Time catcher to solve this if we go back

46:52in time now we’ve learned that decision was wrong do that go and fix it it’s just intriguing to see I guess the net

46:58of it always like because of all of that which is also in you know Justified it’s not like it’s not justified the problem

47:04then moves is then cause to move and that’s where you’re saying well actually there’s a value then in being flexible

47:09because of those situations and that’s where we can meet them but I guess it’s just intriguing more to me to kind of like play it all out so you kind of see

47:16how it all gets to a place in the end it’ll like circumvents into some place

47:21which is then going to be the platform and now the platform I guess has the cost to specifically which is probably value still I’m not saying it’s not

47:27valuable but it’s just I guess it’s almost like saying for the info layer sometimes we do see things in flare where it’s differentiated and it’s

47:33necessary a good example would be like low latency Financial trading like I’ve seen people who are like down to like the nanosecond they need to like because

47:39the faster you get that trade in and out like that will determine the success of your business also like people doing content streaming I’d be like hey come

47:46and use the platform it’s gonna be great for your content streaming and they’re like we need 100 Gig in an hour how’s that going to work not quite like that

47:51on this platform so there are things that even at an infrastructure level are differentiated depending upon the vertical and the business you’re talking

47:57about and that’s kind of what I’m saying like I think a lot of you take some of the educations out and say okay 99 will work it’s like standard infrastructure

48:03and one percent is customized and then you get to the platform and it’s like maybe a certain percentage depending upon the customers we’re talking to

48:09maybe like 30 or 40 is like standard and could just run on like Heroku and then the rest of it needs something more has some custom some elements to it and then

48:15you kind of as you make your way I’m describing a worldly map here by the way actually you make your way kind of close

48:21to the customer so things get more differentiated so we think in the vast majority of organizations like a platform team is necessary if you’re an

48:28organization that’s at scale by that I mean like five teams or more and don’t

48:33need a platform team because you haven’t learned like where your differentiating parts are I’m curious to know how that

48:38works just to look into that and be like I think there may be some efficiencies you could gain by having a platform team because once you get to five teams you

48:44probably do need to start thinking are there common elements on those teams and therefore we could build a platform to support them and just that platform

48:50feature things I couldn’t just get off the shelf somewhere else and almost everywhere I’ve looked we’re working with companies that are like 30 people

48:55and they see value in a platform because they need common things delivered as a service that they can’t get off the

49:01shelf even at 30 people so that’s the kind of I think 30 people upwards that kind of level is where you start seeing value from your platform and the more

49:08teams you have consuming the platform it goes without saying the more teams you have the more you’re like have a force multiplier is the more value you get out

49:14of it yeah perfect I know we kind of running slow and time and I’m kind of like it’s been quite very very interesting because I think I love these

49:21kind of conversations anyway and playing a bit of just listening to other people because I line so like tightly to all

49:27the principles like everything you’ve kind of said but he said makes me almost like deliberately challenge my own view if you see what I mean because you’ll

49:33give me reviews it’s like is this right you know are we doing a good thing here so I was just became Playing devil’s advocate on it because I think it’s

49:39totally necessary but just more for my own like objectivity I suppose it helps always good to challenge our assumptions

49:44because I’m sure five years two years time whatever I’ll reflect on the things I said here and say oh wait that thing was wrong completely I went and

49:50invalidate it so obviously you know our views of the world are formed on a series of assumptions and then we go before hypotheses we go and validate

49:56them so I’m sure there’ll be some things I said today I’ll reflect on and be like what was I talking about millions of

50:01things I do that story in my life I think but if people want to find you how would they find you contact you with

50:07like want to reach out to you and thank you for that chance so be one way or you can go to detail

50:14off again and you can look at the graphics framework I’m also hat of monkeys on Twitter not the people using Twitter that much anymore we should be

50:20using various different things but yeah I had to monkeys on Twitter and on LinkedIn as well so find Colin Humphries on LinkedIn CEO centasso so please do

50:27contact me I’m interested to hear if anything that I’ve said is you find it controversial you don’t agree with it you do agree with it like I’d love to

50:34hear some feedback from the audience about whether or not you think your organization doesn’t need a platform even though you’ve got 10 000 people

50:39that’d be great to hear great hear about those use cases so you get up for being challenged and all of it perfect all right thanks so much as well for being a

50:45pleasure thanks for having me all right thank you thank you thank you