DevOps and MLOps as one platform

welcome to this webinar uh there's a couple of different topics here um starting with devops and mlops in one platform and then if you notice the agenda there is also an AWS uh topic coming out of code commit um and uh what's available from gitlab and uh the neat thing here is uh gitlab has been a really important part uh ethic code is an independent company um we we serve uh devops tools of all types and and partners but uh gitlab has been really good business for us because so many places um have fragmentation in their tool chains for some reason um uh individuals and interactions over process and tools in agile was kind of reinterpreted as uh everybody use your own tools and do things in your in kind of your own Manner and it just resulted in a tremendous amount of fragmentation in tools and gitlab is one of the best Solutions in the world to bring everything together um if you want more decentralized development and more empowerment in your teams then actually I think Consolidated tool chains that provided devop safety net is a wonderful way in order to do that so my name is Mark I am a principal consultant in the devops and Cloud Tower at uh or business area in effic code I've been in Finland for 19 Winters now which the number 19 started a week or two ago and I have with me uh the lovely and talented Yuri aokas hello Yuri yeah hello hello from my my end as well yakas located in Helsinki Finland as you can see I'm an old old it warhorse like I like to call myself I have a long long ssops background I have been setting up Nokia o o store Services actually and and you can imagine how old I am so um I'm really really super interested uh of all all things related to AWS and of course the devop devop SLE is uh super import important for e code so that's that's why I'm here today right and we'll introduce Dan when he comes uh you're gonna have to listen to me a little bit uh just real quick um e code is a biggest devops house in Europe uh we have uh uh 650 people across the land have been doing devops since before it was really a well-known term um we have a a large community around us we just got back from the devops conference in Scandinavia uh we've had people in our conference like Patrick Duo the guy that invented uh devops um we've had a lot of these kinds of uh people in our communities um we do a we're a full stop software shop we do an awful lot of devops work um and we do migrations and we do uh a lot of analysis and assessments and and but basically one of the things that we do at effic code is we don't we don't chop wood we sharpen your axe um and help make sure that you have all the best uh possible software tools and processes available and I'd like to welcome uh what is it number 40 something in the global devops Champions uh Mr Dan Plumley and can I go back a slide I don't seem to be able to go back a slide there we go Dan plumly ladies and gentlemen thank you very much Mark sorry to everybody for being rather unfashionably late there having a little bit of an argument with a zoom update and thank you Mark for for stalling and vamping absolutely perfectly there um yeah just a really quick introduction from me if I may just say hello to all of you at Dan plumbley I'm our gitlab delivery lead in the UK of eff code um it's my responsibility effectively to make sure that people are using gitlab to the best of its and their ability so I do a lot of coaching mentoring support uh sort of ad hoc consultancy as well as a lot of our training um I also just really love talking about gitlab it's sort of my daytoday so this the next hour is nothing unusual for me other than I've got a camera in my face and hopefully that passion comes across to you all today so thank you very much uh thank you very much Mark that is me and you know most importantly you know I I don't and we we aim to educate Inspire and explore at effic code and we don't want anything here to be taken too much as a as a sales pitch or something but we're really excited about what we do and we like to talk about it um we work with uh organizations all over the world uh large very large and small as well as uh we have offices all over Europe and in the United States and uh is it United Kingdom is still Europe it's just not EU right so all over Europe um a couple of quick things this is kind of a preview of the trends document that we're that we're going to send out um there's never been more information available to us uh than ever before on what our uh what kind of metrics are available when we build our software how to measure the value of that software in in production uh being able to get to limit and Empower developers to be able to you know deliver and experiment in production in order understand what initiatives we should invest in um so realtime insights are better than ever before in the world U one of the things we're talking about today is consolidating tool chains uh gitlab being an an excellent choice there but uh really being able to empower people and allow them the psychological safety that's necessary in order to be able to do the type of work with the level of complexity and cognitive load that we have today um it's a really good choice over devel developers spending half their time um maintaining their own tools um something that's super important that's got even more important in the era of AI we call the devops safety net um and this is everything from how you do your test Automation and your your linting and your uh static and dynamic testing um to the ability to roll something back quickly um how you use feature flags do Canary or releases or blue green development um we're also spending a lot of effort on uh Aid driven tooling um all of this information that we've had available for years and you know there have been so many companies building data lakes and just dumping everything in there and with with what's available uh with AI uh today especially being able to process the information that you already have it's never been a better time um and we also do a lot of work in service management we have a a large at laian Wing as well so couple of quick things here and we're going to get to our um agenda so on the agenda um software and model development we're going to talk a little bit about mlops uh to get everybody up on the same page um there's a few exciting things uh that gitlab has to offer some of which I only recently learned about as well which is pretty cool uh then there's some reasons that we're talking about AWS and gitlab um that we're going to get into a while and we're also going to show what some of the opportunities are there and we talk a bit about uh platforms here and a bit about migrations as well um we are monitoring the uh Q&A and um it's private so if you if you send a question not everybody's going to see that and if you don't want us to mention your name we can also um easily respect that as well but uh you will get more out of this if you ask a question and you know please anything goes uh kind of at any level I think that we can handle that today but please do feel free to interact yeah it's like I say safe environment here and uh of course if you feel more comfortable sending questions afterwards uh by all means to do so all right so a little bit about mlops in a nutshell um so traditionally speaking um what uh my experience with mlops I've been in things like uh AI uh recycling robots uh that look at the things that are going by and help you know pick the colored bottles out of a stream of clear pet bottles and things like this but there's a lot of experimentation that traditionally has been done by um AI experts phds um kind of in an ad hoc fashion and um the existence of training pipelines is a relatively new invention like if you compare it to Dev Ops or agile or or modern software development uh uh processes so one of the neat things here is using gitlab um as code repository there are Runners that are available utilizing gpus natively so if you're a a gitlab customer you can get access to essentially mlops runners in order to be able to do the training in in a pipeline uh type of fashion so there are also um built into gitlab um artifact repository and down on the bottom uh on the bottom right in the model deployment there's actually also um gitlab uh has a model registry um so these types of things can be used in order to make sure that you have a structured and repeatable and efficient and extensible approach for mlops machine learning operations so when we were building uh um machine learning models say uh in the early uh 2010s 2015 even um there was a TR a tremendous amount in most shops of manual work that was being done in order to facilitate being able to train a model that um today we have you know things like llms that we're kind of taking for granted uh co-pilots or dual uh gitlab Duo that have kind of been trained on our behalf um but traditionally doing this model creation it not only requires a tremendous amount of work but it wasn't things that were that easy to put into pipelines so gitlab has made that an awful lot easier especially by having uh for example the uh Native uh gitlab GPU Runners and a native gitlab uh model registry so kind of the foundations of mlops so now we made our own slide version of it but as you can see it wasn't quite ready um but there's a couple of interesting points here Yuri yeah yeah just just to mention mention that you can of course like mix and mix and match and then you can like kind of like decide what is the best best uh environment best platform for you to host for example where where do you want to deploy deploy your models where where do you want to score score your models and and then so on so so you can do it do it totally totally in one platform or you can um mix mix and match whatever suits your purpose best all right does gitlab offer a feature store was a question that's actually already came up um yeah I've course done yeah I can jump in there um as far as I'm wearing not in the same way as you know feature stores on other platforms as it currently stands um as Mar mentioned we have things like the model registry so you can store your actual models or or deploy your actual models in in platform I don't think there's a feature store currently um what I would say to that is I I don't like saying no uh because gitlab is constantly sort of evolving changing adding um what I would say is is keep an eye on that because probably at some point as gitlab moves towards this uh sort of MML model uh so not yet but probably will be the short answer probably soon we had another really good question that popped up um which is uh mlops teams are today typically distinct to devops teams it's not only today there's been a lot of Distinction for a long time and I think this has been uh kind of a battle in some organizations where the AI science is done um a little bit in a silo and there's there's not been enough uh devops process devops tooling um devops culture and thinking in the way that mlops has been done so uh the next part of the question is how do you see this tooling consolidation move from git lab working in practice with those organizations with existing in separate teams so really really good question um so the the great thing is that if you the more people that you have working in the same tool chain the higher level of collaboration you're going to have the faster things are going to get fixed um the less uh weight time you're going to have between areas and you know we we look at like developer portals and cataloges quite a lot in the work that we do with our customers and the first thing is always about can we just open up the transparency between the Departments and uh gitlab makes that a lot easier and there's no reason for um mlops and devops to have to be in different tool chains anymore I think that's really cool so thank you for the question keep them coming all right so this is one of the things that we're here to talk about I think this is Yori yeah yeah this is uh one of the main main topics topics of today because I guess we were all shock to notice this kind of like change of AWS strategy that they are actually discontinuing some some services and that is uh quite quite rare occasion ESP especially on the devop side which is kind of like trending trending fast at the moment and one of those like uh the main main things that they informed was that the AWS code committee is going to be kind of like discontinued so they are not taking um any new new um accounts so if you create new account so you cannot cannot access it anymore and they will not bring any any new new features for it they do of course the maintenance so they handle handle the resources and security as as as always but uh um this is this is the moment when you need to start thinking like what what is going to be my my option option for for the future for for um instead of like cod commit so I I think Dan has a really good answer like what could be the aable candidate for that yeah absolutely and this is where Mark and Yuri sort of loosen my leash slightly and allow me to get a little bit biased and a little bit passionate about uh gitlab uh because that would be my answer um sort of taking a step back away from lops just for a moment just for a second um gitlab is the they they call themselves The Complete Dev SEC Ops package or the complete Dev SEC Ops platform right Mark alluded to it earlier when he said that um you know gitlab is all about sort of consolidating reducing our context switching that kind of thing um and that's exactly what gitlab are seeking to do uh in regards to this sort of mlops as well the sort of model Ops uh entity if you like so alongside the really really good work that gitlab are doing with Duo as sort of AI tools they're also moving towards as we've seen and as we will continue to discuss at sort of mlops workflow mlops workflow flow sorry tongue twisted there for me um yes by introducing sort of the module Registries yes by introducing these GPU driven Runners but also as the uh question perfectly alluded to earlier I fully imagine this will carry on um and I will be a little bit biased at talking about gitlab for just a little bit more in regards to why is that a good thing why is that good news why would we want to sort of look at gitlab to to fill this void that platform tools like commit um sort of leave when they are sunsetted um and the best answer for that is not worded it's actually on the next slide it's it's very visual uh the answer to that would be because gitlab had been recognized uh very very recently uh as a leader within the devops platforms by Gartner on their magic quadrants as you can see on the slide there um very much as a leader not just a leader but let's be honest the top leader as you can see on this on the screen there um but based on a number of factors based on the usability of gitlab how powerful it is how good it is at doing what it should do effectively obviously G use far more technical terms than that but I'm aware I don't have hours to be gushing about gitlab here so I'll boil it down just to that um there was also the so that's devops General sort of devops platform quite wide there was also the AI specific Gardner quadrant that they made as well where gitlab were also a leader in that so where I'm really going with this uh and just to wrap up effectively is to say that gitlabs are being recognized at being very good at what they do that will continue um gitlab are very very good at constantly um moving on evolving uh if you like iterating yes by introducing new features but also making sure those features are good they're best at what those features should do so I fully imagine the move or the continued move across to ml will match that as well when we start seeing these extra features the extra the feature stores the uh additional features that will allow us to move these workflows fully into gitlab as well great Yuri H so as you can see our our good friends from uh gitlab also provided some slides for us and um uh here you can definitely see that the gitlab really integrates with a lot of uh a aw your services services and like mentioned before you can really really mix mix and match match your like uh favorite services or favorite uh ways ways of uh working and combining combining tools as a like a really real like tool chain like going going forward forward back and forth for the between the services and uh one thing uh to to mention mention here you can see the code Catalyst here which is kind of like a successor for for AWS cod cod code star which is also like a discontinued uh some some some time ago maybe sakua next slide please uh also of course uh animated animated slide um this kind of like tells you the story of the connection codar connections with the with the gitlab so um you you create this connections and then inside AWS uh either in a console or or in in other other ways you can like connect these Services services to your like gitl uh repositories and and and Runners and um maybe is there some some other way yeah yeah and I have a like a brief example later on in the slides how you do it with the code build but uh this is really really the like the infrastructure uh picture how how it goes with the gitlab so so there is a direct uh in integration which is like secure uh between between these two two platforms yeah next slide please would sorry gentlemen to interrupt would would I be able to jump in only because we've had a really interesting question uh in the Q&A uh from an anonymous attendee um and the question goes uh so gitlab are starting to do model Ops mlops uh they are also offering Duo and building their workflow AI agent y all absolutely true really looking forward to workflows when they come out um but the question is what is gitlabs end game with all of this and I love that question I love the vagueness of it and I love the pointedness of it it's really really good question so thank you for that um if if I may sort of speak for gitlab a little bit I would say probably that their end gain uh so to speak would be to continue being the complete Dev SEC Ops platform right uh I sort of mentioned that during my little Spiel just a moment ago where um gitlab's purpose and for a good long R gitlab's purpose have been you can do everything here you can do the agile portfolio management you can do the security you can do the code uh the the building and the verification the releasing the post- relase monitoring so it almost stands to reason it it goes with gitlabs Mo if you like that they would also incorporate model Ops mlops as well so it is to be the the complete platform the One-Stop shop if you like that's that's one of the you know if we were to go sort of cynical salesman for a second that's one of their selling points is reducing context switching reducing your tools chain down to simple small manageable that isn't their only end game though and this is why I particularly like the question um because it goes so well with the slide that Yuri has just introduced for us as well the integration so it's G laab is not just a One-Stop shop it's also a a pick and mix if you like if I was to sort of extend the analogy to Breaking Point a little bit is that it's just another offering it's something else that you if you're buying a git lab license there's another offering there's something else that you can use gitlab for you you may be using a different tool for your project management you may be using a different tool for your some of your security scans fine you still have the choice to use a different tool for modeling mlops that kind of thing but it's gitlab's way of saying or you could use ours and ours is probably comparable it's probably just as good so that was a really uh vague answer to a relatively vague question but um hopefully the answer was as good as the quest