 I'm Saptan Bharti and we are here in Vancouver, Canada for OpenStack Summit and we have today with us Maksha Daburth. I mean I met you in 2005 in India, we are walking around habitat center and you're talking about space, you're you know travel to space and I mean we have known each other for so long and we are bumping into each other like we last time. NWC multiple other conferences. And last time we met at OpenStack Sydney, OpenStack Summit in Sydney. Yes. So tell us you know how much has changed in these you know six or seven months? So I think the most important headlines would be OpenStack's getting real. So all of the heart air has come out but people still have the hard business problem right which is how do I automate my infrastructure, how do I offer up infrastructure as a service from my own data centers. I'd say AI has come screaming forward we see you know institutions across a very wide range of sectors effectively racing to stand out machine learning capabilities, hire machine learning people and support them effectively. And then IoT has started to sort of really get focused on issues of security, issues of authenticity, integration between the edge and the cloud. Yeah I was at Open IoT Summit and security was almost everybody who you talked to. Yeah. For Zypher OS to everybody they were talking about security. So let's connect some dots first of all. I mean I do know that you have core know when to core for but from OpenStack's perspective when you are here and you meet partners and for IoT what kind of trend do you see? Oh I think the IoT conversation is too far away from the OpenStack conversation for me to be worrying about it here. I would say you know telcos are a place where this comes together. So in a telco meeting we will often be talking to a group that's thinking about cloud infrastructure, age cloud infrastructure and IoT. So in the telco context I'd say IoT and OpenStack kind of meet under the same roof. Right. A few weeks or months ago OpenStack Foundation came up with the white paper around edge computing which is almost all about IoT kind of. So when you look when you talk about IoT from OpenStack's perspective what unique challenges are there because a lot of this may be made at. No I don't think I wouldn't connect those conversations right I wouldn't connect those conversations. So what's interesting about edge computing is that within a large organization of course different groups have a different boundary and that's their edge right so within a large telco you'll have big data centers the edge of that is the beginning of the next transport segment effectively. You'll have a smaller data center which has an edge and all the way out to some router or gateway at the customer premises and each of those teams effectively wants to talk about edge computing. For us you know it's very exciting for us every single reference architecture for edge cloud that's been published by a telco has Ubuntu and has Maz and I think the reason for that is pretty simple. First the one problem that everybody has to solve in edge cloud is physical provisioning because the one thing you can be sure of is that there's no human there right that's you know where's there there is everywhere and so you have to have that bare metal automation that Maz is so good at it's the smallest piece of software that can automate you know from a machine which is switched off to a machine which is running Windows, CentOS or Ubuntu right so Maz is in every reference architecture published so far and Ubuntu is in every reference architecture published so far because of the economics right. People want the latest Linux for their developers because they want to put cutting edge applications there maybe AI applications at the edge but they also know that they're going to have you know hundreds of thousands of servers at the edge and so they want the economics to be really good they won't use VMware they won't use other Linuxes. Well we're talking about Maz when we look at OpenStack you know there used to be three components storage, compute and networking you know so Maz mostly talks about compute right. That's right well it'll also know about all the disks in the machines so it's it's it's sort of a simple version of storage right it'll just tell you you know how many disks are in this machine how many disks are in that machine and it'll let you configure those disks for Windows or configure them for Linux right and then with regard to networking again Maz is only thinking at the level of the rack so it doesn't do all of the complex stuff that OpenStack does with tenancy and software-defined networking all of those things it's really just talking about the switch the switch ports and the servers and the disks. But two other component storage and networking is also critical and canonical works with you know all those telcos as well so how are you also planning to bring some story solutions or you're planning to engage with the NFB communities also from canonical perspective or not? I think there are two obvious storage stories that go everywhere one is directly attached storage which is basically just the disks in the machine that's hosting the app or the VM and the other is Ceph as a distributed store now we do see appetite there are people who are buying proprietary software-defined storage solutions to integrate into OpenStack Kubernetes and other things at this stage I wouldn't say that there's any of those that strikes me as being a breakout success but clearly there are some gaps in the market because we do see companies you know selling to our customers as part of canonical OpenStack or sort of edge cloud distributions. I'll just talk about OpenStack one more time before we move to it. Last time when we met at Open Sydney the focus from the foundation was more on collaboration and composability what has been the theme this year? Well the foundation has generalized its mission from OpenStack to Open Infrastructure and they've introduced a couple of other projects alongside OpenStack, Carter containers, it's really a sort of a next-generation KVM from Intel where I think there's some really good research and development I mean there's some really good thinking about virtualization right which I think is actually useful for OpenStack generally but they've positioned it as a container security mechanism and then Zool which I haven't really looked at so I can't speak to. You know this summit for me has been very interesting you know that I caused some controversy by putting a spotlight on the economics of ownership but I think it's very clear to me that everybody who's here at the OpenStack Summit is here because they feel they're responsible for a data center and they want that data center to have OpenStack in it and they want that OpenStack to work and they want it to cost a reasonable amount of money. We can't ignore economics the public cloud is great and the public cloud makes data centers a simple economic proposition so that's why I took such a strong focus on the money right for people to want OpenStack they have to believe that it's going to come in at a price that makes sense in a cloud world and so I was simply showing that Canonical delivers OpenStack we deliver it within two weeks guaranteed we deliver it with hardware partners so we really simplify the acquisition of the hardware the reference specification all of those things we will operate it and when we're operating it you have the exact line of sight on pricing right you know exactly what OpenStack is going to cost you it's half of what you might pay for kind of traditional proprietary enterprise solutions and that means it's competitive with the public cloud I'm not trying to say people shouldn't use the public cloud actually they should use the public cloud right you want to own a house and you want to rent some space occasionally right or sometimes if you need a conference center you want to rent the conference center right public clouds are great private data centers are great if they're economical and automated and when you said open infrastructure where does Canonical fit into that new theme of OpenStack the vision for Ubuntu is that it delivers the best open source software and it delivers it in a format which is enterprise grade upgrade friendly developer friendly right with it with a clean and cost-effective support structure operations set of services around it right so whatever open source emerges from the open infrastructure world or from the container world or from everywhere else in open source right our job is to make sure that you know what comes out of out of research and development or community development is consumable by enterprises in a way that is super predictable and to do that more cost-effectively than anybody's ever done it before initially you also touched upon the hype cycle of is over nowadays we often hear you know a Kubernetes is everything or containers are the days of open stack and number what do you have to say about that is it like true or they will I don't see open stack and Kubernetes competing with each other I mean we we have an offering where we say we will build an open stack and fully manage it and then you can have as much Kubernetes as your different dev teams want on top of that open stack and it's all included one price so those two things don't compete with each other at all another example is for AI say you say you've bought a bunch of racks and you filled them with GP GPUs and now you have lots of different teams in the business that you want to be using those GP GPUs open stack is a brilliant solution you redeploy open stack we create instance types that have GP GPUs teams then create their own Kubernetes on top of the VMs the performance of the AI code inside those VMs isn't really governed by the virtualization right the performance is all on the GP GPUs right but you have very good security and isolation and sharing of those GPUs between different tenants effectively in the building so to me open stack and Kubernetes go together really elegantly you see the same thing on the public cloud right the Azure Kubernetes service on top of Azure you have the Google Kubernetes engine on top of the Google cloud right so I don't know why everybody's all kind of the reason I asked you this question is because I want to reiterate the message you know that there is no conflict as such no sometimes you have mentioned being mentioning AI a lot of time I will go to AI at the end but I'll just touch upon IoT one more time before the use cases of IoTs are like industrial IoT is Nassio consumer IoT is totally different so various canonicals focus remember we were a gateway to the open source platform right and so I didn't want us to spend all of our attention on self-driving cars right the self-driving cars generally run Ubuntu but I didn't you know that's work for Audi and Volvo and and Google and Uber right I want to make sure that that if they use Ubuntu they get a highly secure platform where they can do updates three times a day right with perfect reliability yes right so we've invested a lot in security we've invested a lot in new update mechanisms and software delivery mechanisms to Ubuntu but our product is just Ubuntu like we're not we're not offering a self-driving car solution we're offering the most reliable Linux for self-driving cars right we're not offering a drone solution we're offering the most reliable Linux for a software defined drone platform we're offering the most reliable Linux for an intelligent camera platform we're offering the most reliable Linux for you know a refrigerator platform wherever people want to go and innovate that innovation looks like something physical a little server motherboard Ubuntu Ubuntu core if they're doing it at scale and in production right and then a set of snaps that represent the unique capabilities of that piece of IoT right the motherboard is standard Ubuntu core is standard we take responsibility for the kernel the OS effectively and all the security associated with that we take responsibility for isolating the applications from each other but they take responsibility for their apps right when you look at IoT there are I see three components one is hardware one is of course the OS that is an application that people run and in terms of security all three play a big role recently Microsoft announced a sphere OS which is running Linux kernel but they also have MCU you know so they have some influence over the hardware as well so when you talk about IoT what kind of influence canonical and having the hardware label to ensure that that those two layers are also secure so so first I kind of just wanted to celebrate the moment of Microsoft sphere right at first I think it's a it's a it's a great moment for Linux and it's a great moment for Microsoft right second it's a great moment for Ubuntu core because I read the the sphere white paper which is published in 2018 by Microsoft it's super credible forward-looking technology company and there are seven key ingredients that they identify if you look at our design documents for Ubuntu core we identified all of those things and we built it as free software over the last six years so what's kind of yeah it's kind of fantastic for me is you know we always talk about how free software is always kind of playing catch up to the big guys and proprietary guys but look at what we've done with the Ubuntu core right where we started with six years ahead right and we've delivered this completely open platform right essentially a free software IoT platform that meets every single one of the tests that that the Microsoft guys put out in their white paper that's amazing all right so I'm super proud of that so I'm asking is it was there an influence from canonical or two or it's like independently they were no no I'm not suggesting I'm not suggesting that and we certainly never had any conversations right I think small people come to come to conclusions but I think that's a real deep validation of of our kind of thinking deeply about the next wave of problems and then investing in in things that maybe not everybody understands so they can easily become controversial you know how things can get controversial in the Linux and I try to avoid that yeah right and yet if we don't do that we can't be there at the table when the game really unfolds right so I mean I think our Ubuntu or Unity has been ahead of the time a couple of times but what then goes wrong you know that you know we don't kind of you know succeed in delivering you know the I mean we have great ideas you know as you said you know you're ahead of Microsoft in that sense oh look I think you need you need a bunch of things to come together to be successful you need to have the right idea right you need to be committed to that you know nothing useful happens you know overnight it takes 10 years to make an overnight success right I think you need to have the trust and support of the people around you and I think that's a little bit difficult because I think that the open source community in the Linux community is very quick to rush to judgment on new things and very quick to get tribal like if that's a new thing from x oh then it must be a terrible thing right we haven't even looked at x haven't looked at the source code haven't you know what I mean but people get very emotional about things that don't matter like which company it came from good example of this is MySQL from Oracle right it's still MySQL it's still open source they're still investing a ton in it right it's great right but suddenly it's easy to troll people into getting angry about that because it's Oracle that's crazy that's crazy right so I think you you you need to have the right idea you need to have the support you need to get given the rope and the community does need to sort of get used to trusting people to go chase ideas like that chase a dream you need to execute you know if the code is shit then you're gonna fail right right and and you know if I look at Unity for example it was the right idea we had some we had some stress with the community there was also problems in the code right you know we we we were doing a lot of things too quickly and and and so then we took too long to get it to a position of real quality and I you know I'll take responsibility for that right you also need to have a business model for it exactly we forget about that in open source but the truth is you know people will have to eat and you know they will disappear into other proprietary worlds unless we can find good ways to feed them and so again when ideas have been put to the open source community it's healthy for the open source community to respect the idea that there's business going on that's part of why I stood up here on stage and said look focus on the economics of the data center right you can't just kumbaya and says all wonderful that we're here together again in Vancouver talking about a bunch of these shiny projects which might be useful in five years time if you haven't yet solved the economics of the data center right after seven years so I think all of those things have to happen you know when you when you get all of those things together it's great and when you don't you learn something right right right one of the one of the key points of your keynote actually was the demo that you gave at the end but it got kind of buried in the controversy do you like to talk about that or should we switch to AI sure so what what I demoed was work between Google canonical and other companies to standardize the sort of open source workflow for machine learning AI and we and I wanted to show how that helps people essentially empower a set of developers to work on AI applications effectively which they can use in the data center on open stack or bare metal so that's basically I was showing canonicals open stack canonicals kubernetes and then kubeflow on top of that or they can take those apps and take them to the Google cloud and there you've got the Google cloud Google kubernetes which uses Ubuntu effectively so we can have exactly the same stack and kubeflow on top of that again which uses Ubuntu so we have this portable layer between your data center and the and the public cloud and then I showed something really cool which is microkates exactly it's a it's a one command install of upstream kubernetes it's super light it works on fedora works on open suzer it works on Ubuntu works on debian right so any of the major Linux distributions and any of the minor distributions that's like like the specialist distributions that's keen to have snaps it's dead easy to enable snaps now it's all upstream so snap install microkates and you then get a a single node kubernetes the latest version of kubernetes and kubeflow can sit on top of that so now from your laptop to the rack to the cloud you have a beautiful kind of developer cycle and then you could take your AI application you can take your model that you've trained turn that into a snap and ship it to iot okay before we wrap up you're here at the conference you have met a lot I mean you meet your partners either way that you don't have to be at a conference what are the big challenges problem that you think that they are facing that the the community is still now solving or looking into that should be solved I think the the biggest challenge with OpenStack is the complexity and diversity of options okay that's long been portrayed as a strength for OpenStack but I think what we're doing is we're saying look here is a set of free-tested reference architectures for OpenStack which are in production at telcos at banks at media companies at retail companies if you use those you are benefiting from all of the knowledge and all of the mistakes at all of those other companies and so it dramatically reduces the number of choices people have to make they still have choice of certain sdns for example they still have choice of some software defined storage options it's not a lockdown architecture but it is essentially a tighter sort of universe of options which are known to work known to be upgradable known to work efficiently so that's really been our kind of ruthless focus over the last year is take all of that complexity of OpenStack distill it down to the choices that really matter and then do everything we can to make it really easy for companies to go from zero to one rack of OpenStack at a reasonable price we put up this three hundred thousand dollar offer right which is one rack of great super micro kit plus a full year of managed OpenStack from Canonical so essentially you need no OpenStack skills up front you just get an OpenStack cloud now your developers can start accessing that and you can grow it as much as you need and then when it makes sense you can train your people to take over the operations from Canonical right right but on the keynote there Markolier he said you know the OpenStack is you know diversifying with the cloud it's kind of contradicts you know a kind of well you know in a in a in a community you're going to have different right that's true yeah so so clearly each of the members of the community is entitled to focus on the things they're focused on what Canonical is focused on is time to value and the cost of the OpenStack so how can we deliver OpenStack faster how can we deliver it more cost effectively right i talked about our commitment that we will deliver OpenStack anywhere in the world in two weeks yeah right and i talked about the fact that we will do it at a fixed price which fits in the budget for any department head right so any department and any company can get OpenStack in two weeks they'll then have time to figure out if that really works for them and how much they want to invest in having their own version of that or you know blinging it up right but what they what they what they don't have to do is spend months and months and months kind of studying and figuring out and making a bunch of choices that they don't really understand right because we'll put them on rails and last question then this final when i talk about Kubernetes Canonical is there when i talk about OpenStack Canonical is servers what is Canonical to do what kind of company is it well we we focused on the public cloud the data center and the IoT and so it'd be natural at any place where open source is being used as the first port of call for the data center public cloud or IoT that Canonical is going to be there because people associate us through Ubuntu they associate us with having the the best reference distribution of the best current open source software thank you Mark so much it was great talking to you again