 Okay, this is a panel discussion about public Cloud Foundry platforms in Europe, and it's really intended to be useful if you are thinking about using public Cloud Foundry, public Cloud Foundry, or you are operating your own Cloud Foundry and you would like to learn from some of the people that do it at scale. So before we start, can you put up your hands if you are already using y public Cloud Foundry Paz. Take note, panellist, so you can respond appropriately to the audience. Can you put up your hand if you don't use one, but you operate your own Cloud Foundry instead? I'm going to ask the panellist to introduce themselves, but before that I'm just going to make a special mention for the ones who couldn't make it. Holly Cummings from Brumix Garage, sadly couldn't be here. Bridget from Pivotal and Rebecca from Noe Anthony from CenturyLink couldn't come, so look forward to having them on a future panel. My name's Dan Young. I work at a Cloud Foundry consultancy called Engineer Better, and I have used a number of different public Cloud Foundry providers, so I have opinions of my own. I will now let the panel introduce themselves. Hello, I'm Jonathan Matthews. Despite what it might say in your apps and your schedule, I do not represent any large UK retailer here today. I'm an independent Cloud Foundry integrator. I've used all of the people on this panel, so I've used products provided by all of the people on this panel today and have opinions about all of them and some of the people on this panel as well, but I'm sure we'll not come on to those today. I am, if you like, the slight voice of a customer in this scenario. Thank you. I'm Julian Fisher, CEO of any9s, and we started as a platform and now also do consultancy around Cloud Foundry. I'm Colin Humphries. I'm CTO for Cloud at Pivotal. We don't have a public European pass at the moment, but we're giving it a lot of thought. We do have a public pass that operates out of the US. My name is Mark Oxtraser. I'm running the Cloud Development in Swisscom. Swisscom is a Swiss service provider, and we run a public offering of Cloud Foundry, as well as a couple of enterprise offerings around that. Hi, I'm Dominic Harrys. I work at the Bluemix garage in London for IBM, and I'm slightly on the user side, I guess, of the public cloud using Bluemix with customers to run projects. If I was to look at... My first question is going to be operations-related, I think. If I was to look at all the different API versions, CloudControl API versions, if you go to cfapi.buemix.net, somebody's written an app that shows you these, they're all different. That would imply that there may be some challenges in deploying Cloud Foundry at scale. Why are they different? Who would like to have the first question? Why is yours different? You mean the platform? The versions of the APIs that are being used on your platform. If you compare that with Pivotal and with Bluemix, they're all different. They're all slightly different versions. The only thing that's different on our platform is the data services. Despite of that, it's just the regular Cloud Foundry. That's the answer to the question. I was just looking... What challenges might you have when deploying Cloud Foundry means that you can't always just run the latest version? You all appear to have different cadences of deployment and upgrade. I'm wondering if that possibly, as a consumer, I'd be interested in understanding why you have the particular cadence that you have, providers. Oh, I see. One thing I'd like to see around the Cloud Foundry community is to organise, let's say, a delivery cadence and maybe a release management that makes it easier to do that, because as an offering based on open source Cloud Foundry, you have to manage that yourself. Of course, you have to provide a CI pipeline, go through testing all the components you want to apply to a certain version and see whether it works with all the other components you might have other providers don't have. For example, our billing system that takes usage data and translates it into something that we can generate invoices from. So, yes, that's something that would be desirable. Of course, it is a certain burden because you have to organise that and consumes resources that could actually be shared with other Cloud Foundry providers as well. Pivotal web services runs the latest version, the bleeding edge. I haven't looked at the other APIs. Last time I did check, Pivotal web services was ahead of all the other providers. We're running the latest versions. We test it before it goes to Pivotal web services, but then it runs as a service in Pivotal web services before it goes into PCF, which is our product that you install on-prem. You have that additional layer of testing with thousands of users before it becomes part of the on-prem product. Yes, so we do it almost the same. We run roughly two to four weeks behind the official release, which is roughly second or third after PWAsco's life, and we try to be very, very at-hawke. For various reasons, there are CVEs, there are security issues, so you need to be very much on the release, which is heavy for the users, because sometimes API switches, you just need to switch, as Julie mentioned, so you have challenges there, but I think we really focused on having a system to really upgrade very fast, because I think this is our duty to be on track. It's an obvious product. We will find bugs, we will find security issues. We need to be very, very much on track, and this is very focused on. So Bluemix is massive. I think it's got more users than any other public cloud platform, so we have to do a bunch of scalability. 1,000 users a week? There's a lot of scalability testing and that kind of thing, and I know there's a bunch of custom security stuff and other integration things that go in. The other interesting thing about Bluemix is that we have single tenant environments and local environments that we manage for our customers, and so then they get a bit of choice about when they upgrade, so there's a specific event or a specific load that they need at a certain time of year. They can postpone the upgrade just to make sure everything stays. So there's a window that customers get to choose where they upgrade in. OK. I don't need to keep running over that, because I've got my turn here, haven't I? So what feature do you not have that you wish you had as a provider in Cloud Foundry? What is it that you... It might be something that is coming down the road, but isn't here yet, or something that is not even conceived of yet. What would completely change your ability to operate at that sort of scale? Or would help the business? I think... So this is something that's coming. The ability to have seamless deployment to multiple availability zones. So we have patterns to do that, which require a bit of manual work. We have some services in beta. There's a CDN service in beta and that kind of thing, which helped developers to do that. But having kind of a single CF push and magically everything just works would be great. Obviously it's very difficult. I think, as we talk about public Cloud Foundries, I think it's also important to see the essence. Why do we need public providers in Europe? Why is Cloud Foundry worthwhile to run publicly? And I think it's important for our customers to move workload around to select which country they want to save the data if they need to. And this is obviously why we build it up in Switzerland. There are certain customers who want to have it there. And what is missing there and what I think we can really leverage ourselves as a group. I mean, we are competitors to a certain kind. But I think the next level we should aim and what I'm really missing is that we heavily share our services and our service marketplaces. That we heavily go forward and say okay, you can get all the whatever services in this cloud and in this cloud and we just share it together to give the customer the full experience of a public mesh of Cloud Foundries in Europe. I think this is the biggest miss I see. I think the biggest feature I think coming down the pipeline is going to be the isolation segments. If somebody in the talk earlier on asked me to run apps on a separate set of infrastructure so I could see for public cloud providers you could have a customer who is using Cloud Foundry quite happily and then has some apps that need to be essentially on their own set of infrastructure. So you could then allow a customer to burst out to their own infrastructure for a particular app. So it gives you the isolation of a set of runtime components but it means that the customer doesn't have to pay for all of the control plane and it should benefit of being able to isolate particular work loads as required. So I think that's going to be useful for public pass providers. So I totally agree that multi-data centre awareness is something that should be part of Cloud Foundry at some day as it is a motivation for many Cloud Foundry users or adopters to go for multi-region strategies. That's true. I also agree to Marker's opinion that data services are important to have a certain set of common data services so that users can develop against the Cloud Foundry platform and that there's ensured that you can just take this application and run it somewhere else as well. And yes, that's basically my thoughts on this. Could I have a go at answering this question from a... Well, I was about to ask you what's the best thing and the worst thing from the user perspective about using public Cloud Foundry? Could I touch on that question previously, which is the biggest missing thing as a consumer of public CF in Europe is we have, I think, four providers here, only two of whom are actually technically if you want to get your lawyers involved in the EU. I believe to what it's called, you're not strictly in the EU from a... How does it look like the EU? It's like Europe. The world should be very specific. So... And it's called... I mean, let's talk about Britain. Yes, yes. That's another one that I don't really want to anticipate to taking a good couple of years off at the point that that happens. It's only Julian who's left then. So long story short, from a consumer perspective, the thing I want is more. More EU providers who I can point out and go, this is definitely in scope for things like PCI, this is definitely in scope for personal data, things where I don't have to cross any potentially what's the difference between EU and Europe boundaries, things like that. And then I've forgotten the question you actually asked me. I said what was the best, what is the best thing and the worst thing about depending on a public cloud boundary provider for your, instead of operating your own... Simply the best thing is I'm not running any service. That I'm sure everyone who consumes public CF is aware of that as being a huge win. The worst thing I think is mitigating the difference in support and the difference in transparency that you get from different providers in the case of problems happening. Perhaps not intentional difference in transparency but just the effect of how their support works and the way that they communicate that with the public. That has been a significant issue for the organisations that I work with. Can I jump back to a previous point? I realise we sound like we're in furious agreement at the moment on this panel. I just noticed something I wanted to pick up on. Maybe a little bit contentious. So we're talking about services, the different services on the providers. Julian and Marco mentioned how they think we should have a common range of services. And also pivotal sales pitch. The pivotal services you can deploy from PCF, you can deploy them to any IS provider that you want so you have this kind of data service down wherever you want. But I wouldn't want to sit on this panel and look to use IBM of trying to differentiate by their services and lock you into IBM's paths by providing the services that are behind there. I wouldn't want to say that. But if that was said, how would IBM respond to that? So I would say, I don't know what you're talking about, if you're talking about Cloudant, if you're talking about Cloudant, that runs on AWS and Azure. If you're talking about compose.io, that runs on, not today, but stay tuned. I did want to touch on the motivations of the different providers for being in the market because I think it helps users to understand that so they can make better decisions. So I think Colin might have alluded there to some other motivations allegedly that IBM would have which is, you know, we've got this big marketplace of middleware services and the past is kind of a way of reaching that. But Pivotal obviously are using the very latest code in PWS and John, what did we discover yesterday in the training? Things were a little bit wobbly of these sometimes? Yes, I can absolutely see the... Can I keep it here? Yes, I can absolutely see the benefits of doing dog fooding the product on before big enterprise customers get the code coming down, but we certainly couldn't see any signs of the slight wobblyness yesterday being recognised anywhere. Although to be fair that was a relative outlier. I'm purely talking about PWS there. We don't generally see that as a problem but we definitely did see we didn't get information coming forward about that that I'm sure you would get if you were paying big, big enterprise bucks but it was definitely interesting an interesting day yesterday given the context that we were sitting in. It was... I can... I wasn't there the training yesterday. Obviously PWS as I mentioned earlier does run the latest and greatest code. We put it through extensive testing via concourse. Heavily tested before it goes out but we do have... bugs will emerge. I can say I've been a pivotal for seven months prior to that I used PWS extensively as a customer and in my experience it was the most stable public cloud foundry. We're looking potentially at bringing it to Europe running PWS in Europe but as I said earlier that means if you're paying for PCF you get something that's far more heavily tested has gone through far more in terms of has all the security fixes everything else we can possibly get into there to make it as good as possible. So obviously as PWS PCF is our focus PWS is something we use to make PCF better. So yeah the message there is that it's not it's not meant to be a profit making venture. It's a very useful service but it also helps pivotal actually develop their commercial products. It gives you a great try before you buy and it's fantastic for really all the code we ship we run as a service before you run it as a service but our focus is PCF. So one of the customers that I've been working with had an incredible statistic where they said it was their Christmas food ordering website and they had moved from a situation where they were spending $1,000 on hardware just for that short three month period of the year and then they'd moved to a public cloud found your provider in Europe and they're now spending $200. So my question would be how is anyone making any money out of that? What kind of scale do you need to be running at for that to be a viable business? So I've been through the belly of darkness with public offering at any nines. So we started at a listed v-sphere cluster and moved to a self-hosted open stack which was the most horrific thing I've ever done. And calculating prices against Heroku the largest competitor two challenges are when offering a cloud found platform with no domain specific differentiator such as being an IoT platform or something which is first when Heroku came up it was seven years ago I think then there was a wave or there's this something new everybody looked at it just like the hype around Docker is today and you will not get this second wave so you have to find your self a marketing that has more you know other points of attraction and the second is infrastructure so infrastructure in general is a very hard topic and our value proposition used to be 100% European currently our target sits on AWS because we were unable to find a trustworthy infrastructure provider being entirely operated in Europe not being held by a US holding company that offers us infrastructure as a service so we can build a platform on top of it. The calculation from a cost perspective with a self-hosted open stack allowed us to be competitive with Heroku from day one even at a very small scale so from an economical point of view I'd say it is possible that let's say smaller data centers use infrastructure as a service and put a cloud from you on top of it and make a viable business out of it in case they are entirely focused on that we've been distracted by consulting business also and if they manage to get the infrastructure stable maybe chewing on your first point we are happy to help you with the infrastructure we are a non-U country but in Europe like every UK provider so maybe let me know we also run an open stack which obviously helps us bring the cost down it was not easy a lot of people here in the room probably think of these days right now it was hard but I think it's worthwhile the public offering was especially in the beginning it was clearly intended to be a business card to show that we are a cloud provider that we have passed in the market and also to push the cloud foundry message our main market was at the enterprise offerings where we have dedicated cloud foundries fully virtual private where we encapsulate network and all that this was our target what we saw in the last six months is that our public environment really grew and we are on a stage where we said it's way bigger than we ever thought it will be and also the revenue economics really work and we also see quite a lot of non-swiss customers you know for whatever reason put their data in non-U countries but it's fine we're happy to get it I'd say the kind of cost per hour is one cost that you'll have on public a lot of IBM's customers obviously enter into more comprehensive agreements around what they're providing around support, around services all that kind of thing which I think allows that basic cost per hour to be very competitive but also partly as a calling card to maybe a single tenant environment or that kind of thing but also as just something which opens you up to services that may be chargeable or support that may be chargeable as well So of those 20,000 new users a week what proportion of those are startups because obviously Blue Mix is quite focused on startups That's a good question and if I had the numbers I don't even know I'd be able to tell you but we work in the biggest I think the biggest startup co-location environment in the world and we work in Moorgate and we've had some really great successes with startups coming along using the platform it's very easy to get started with the built-in DevOps and things like the Watson services which allow you to spend pennies for stuff that you would have had to have a massive contract with a big supplier for before I guess the same question to the others the other providers here are you seeing more enterprises starting to say we don't need to run our own cloud boundaries? So we see both but we see more enterprise and we see actually more enterprise developers which are kind of stick with the internal IT situation and then move into our cloud so for instance two months after we launched our public cloud one CEO of a bank in Switzerland called me to ask me if we can shut down the organization which some of his team did that we should stop his team pushing apps to our cloud because they're not allowed to this was a very nice conversation I should have recorded it but I think we have quite a lot of organizations which are really big organizations where some developers really try to have an easy way still deployance with some but have a very easy way to use cloud so we see but it's I would say half half right I think we found it quite interesting how a lot of our big enterprise customers they have VMware or OpenStack on-prem they buy PCF to deploy onto it and then interestingly quite often they either hit capacity limits with what they can have on-prem or some of the reason they can't to scale anymore and then they'll just put Cloud Foundry on Amazon they'll just scale it out and that's where we see huge amounts of value where all of a sudden they were used to hitting some kind of a capacity limit and now they can just be like oh we'll put PCF on AWS we'll scale that out the apps run exactly the same the data services run exactly the same so we've removed that friction between being able to take workloads from one infrastructure and put it down on another one we haven't seen a lot of people to running their own Cloud Foundry on AWS to then starting to use a public cloud provider as well I haven't seen a huge amount of that but maybe that's because that's not really been our focus I think it's more gone the other way people will try out Cloud Foundry on PWS enjoy it and then like to install PCF either on-prem or on AWS so I'd say that the size of Cloud Foundry makes sense to be on-premise you know how big should a company be so that it makes sense to use your own Cloud Foundry I'd say that it's it's surprisingly small so you can make use of a Cloud Foundry in a certain circumstances being dedicated however the cloud is about sharing so when you look at at Amazon at a certain time they just open up their infrastructure to be used by others so what's the big value in using Amazon has a let's say unlimited capacity from our perspective as a small organisation I'm running a 50 people company here like the way the smallest company represented on the board I guess so for us Amazon is unbelievably big where our infrastructure had around let's say 2 terabyte of RAM so when a customer own boards a small platform your question is how much of elasticity do you actually need and the same thing is when you run dedicated cloud phone within your data centres how big is your infrastructure how much of that infrastructure is taken by your Cloud Foundry and how much elasticity do you need within the runtime and within the infrastructure part of your infrastructure so with that the question it could be that you have to order new hardware because you hit a boundary because there was user growth and you did it unexpectedly so in that case on Amazon you'd possibly be able to just use more resources because they just have more free space and that's a calculation everybody has to do are you big enough to make use and to benefit from the platform within that boundaries of your data centre or your core location is there any audience questions before we reach the end I should really check who's got burning questions for the panel because we've got the microphone anyone at all Alex do you have one you did safe harbour yeah so I think the question was does pivotal plan to offer the services from PCF on PWS is that correct yes we do we're looking at that we're making them incrementally more and more available to customers so we're very much doing that but that's taken a while I admit but that's definitely happening now I think the question was are we also going to run the tiles from other companies on PWS I think that's maybe happening but I don't want to make a commitment right now we're agile it's certainly something we'll learn from running our other PCF tiles on PWS and then if that's successful we'll then look maybe with our partners to bring their offerings as well that's a great question and it does play towards the title of the tour which is what can we do next how can we really improve this marketplace maybe just touching on what Marco said that positive sun game aspect of working together as a community to improve things I think that more visibility about how one can transparently or easily migrate across different public providers would be from a consumer's perspective really useful there are currently a couple of some challenges that we face so I think that would probably be a good thing for the whole community of public consumers to be able to say we can take a bit of service here and a bit of service there and not have to hugely care about the distinction between them outside of the services later I totally agree so taking the perspective of the Cloud Foundry end user I want to set up as a Cloud Foundry CLI on my computer I want to set the target, deploy an app and I want to do the same thing by just changing the target that's my desire and I don't want to care about the versions or whatever when I have a manifest file specifying the dependency tree of my app I don't want to worry about anything else that's the main motivation why I use a public pass and currently there's so much heterogeneity that there's so many differences between the platforms that this isn't possible right away and I think that is improved that will be a benefit for the end user I'm probably going to answer this subtly differently so what I would love to see happen within a year would be some large European companies take part in Cloud Foundry dojos have more core committers to Cloud Foundry from Europe and bring in platinum members gold members to the foundation of Cloud Foundry from European companies I know we've got a few, I'd love to see more of them I'd love to see far more other companies getting involved far more core committers Cloud Foundry is ours so let's invest more in it as people, let's put more people onto it let's just help it grow I think also it goes in the direction of growth but I think we just need more providers as well I know a couple of providers which will probably one will probably launch already this year and maybe one other I know next year, early next year so this is cool but I think we need more so it's a good business to be in if there's anyone who wants to start at Cloud Foundry you recommend it I think we should make it very easy for service providers in Europe which are dedicated in countries to launch something like Cloud Foundry and this also goes in the direction how small does Cloud Foundry need to be to run it by yourself it can be very small but it's very hard to run it it's very hard to run it you need very good engineers you need people who know how the upstream works you need very fast cycles and security updates you need to have your hands on that stuff and I think this is very important so size is not the only thing PCF helps with all of that in case you were wondering okay that's it, we're finished unless there's any more questions I think we're probably keeping everyone from their coffee so Dom did you have something else to add? I would just say along a similar line I think the seamless transition so that if you have some stuff ready in AWS or some stuff you want in public cloud some stuff you want in a single tenant cloud or on premise just a good story for how you can easily connect those things and work with what people already have okay great thank you very much to our panellists