Cloud Native Computing Foundation KubeCon + CloudNativeCon North America 2019 (San Diego), 22 Nov 2019

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Intro + Deep Dive - Provider IBM Cloud - Sahdev P. Zala, Brad Topol, Richard Theis & Ian Watts

Description

Join us for Kubernetes Forums Seoul, Sydney, Bengaluru and Delhi - learn more at kubecon.io

Don't miss KubeCon + CloudNativeCon 2020 events in Amsterdam March 30 - April 2, Shanghai July 28-30 and Boston November 17-20! Learn more at kubecon.io. The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy, and all of the other CNCF-hosted projects

Intro + Deep Dive - Provider IBM Cloud - Sahdev P. Zala, IBM; Brad Topol, IBM; Richard Theis, IBM; & Ian Watts, IBM

Recently the Kubernetes community reached an important milestone of moving individual cloud provider SIGs as sub-projects under the Cloud Provider SIG. The SIG IBM Cloud is now a sub-project called Provider IBM Cloud. In this session, the Provider IBM Cloud leads will discuss these recent developments and brief about the activities of the sub-project. This will be followed by a deep dive into the Kubernetes Cluster API Provider IBM Cloud and the recent development activities. We will then provide the current status of the IBM Cloud Provider. Everyone --active contributors, new contributors, and conference attendees -- with interest in the Provider IBM Cloud sub-project are welcome to attend. During the session, the project leads will be available all the time for discussions.

https://sched.co/Uakh

A

Welcome to the Florida IBM cloud introduction and deep dive. My name is Serizawa. I am here with my co. Speakers brought opal fresh artists and ian what.

A

So for today's session it's a it's a marathon, a team, eight sessions. We have combined both introduction and dip-dye together. So we a lots of time. Please stick around. We can go to the presentation and then we have time to have some discussions around your questions and whatnot agenda.

A

I'll just walk over walk you through the overview of the project, will walk you through IBM cloud. You know this is an IBM cloud provider session. So we're gonna walk you through the IBM cloud. Then we have deep dive around the activities. We do we'll talk about the IBM cloud provider, I'm gonna, do the cluster API provided IBM cloud and then we're gonna have a discussions with you. We have some topics in the discussion, but anything from you alright.

A

So before we go talk about the provider IBM cloud, let's talk about cloud for your seek, not sure if you're familiar with the cloud provider special interest group, but the the seek was created over a year back last year, early in the last year and was created with a few things in mind. You know the first one is to ensure that cuban it is ecosystem evolve in a way that is neutral to all cloud providers right. So, as you know, there are several cloud providers in the Cubans areas and there will be many more right.

A

Even you think that a lot of companies moving in house to their infrastructure using Cuban IDs- they are you know, also a part of that journey, and the seek is also making sure that you know the the the Cuban is requested of evil in a way that provides a consistent, high-quality user experience across the providers.

A

So what you this? The cloud partners see. You know it defines the cloud provider interface and provides mechanism to initialize a crowd forward implementation. You know there is a controller for routes and no and for services yeah, you know so we're gonna, deep dive on that. A researcher will cover something called CCM. The architecture for this interface before interest is there's a link and get up calm. Cuban is cloud provider. You can find kinda tons of information there.

A

So what happened in this area and the cloud part is seek, like I said he was created with couple or few different purpose. One of the purpose was to combine different cloud 46. So you know until like few months back, we had we had a special interest group in you know different cloud products. Oh yeah, there was like a sig called sig AWS. There was a sequel seek as you're seeing.

A

Out and seeing OpenStack and sync VMware, so we have different SIG's for this major cloud providers and, like I, said there are many more cloud forests out there, but with this cloud were sick, you know the community thought it's a good idea that we bring all these six into into like one umbrella into one club, one unto one sink so that there's a bit of collaboration across the you know the provider. Members and you know we can do our job better right.

A

So what happened after the last cube core I'm, not sure if you, if any of you were at the cube, can Barcelona, but we had good discussion there and after after that, cube corn, we have, you, know, created some open documentation. We work with the steering committee, all the different cichlids in the cloud for idea. You know they came to a conclusion that hey, let's just move under the cloud quarter, see as an working group.

A

Alright are a sub-project, so IBM cloud specialist group which was called formerly it's now called provider, IBM cloud and it's a sub project or working group of the cloud for their sake. Alright, so what we do under this working group or under the sub project I said all these cloud products has the you know. The similar purpose of you know how you can will deploy.

A

You know support the Cuban. It is on the particular clock right. So, if your is an IBM, only clouds private cloud, if you are, you know using different cloud providers right. How do you, you know, create your cluster on that infrastructure? Right? What are the different ways you can use to leverage the resources? Particular cloud provider provides okay from IBM side. We have public clouds, we have private clouds.

A

The both are participate in the cnc of certified Kuban, Ares conformance program, not sure how many of you familiar I'm sure a lot of you, but this both clouds from our side are certified and there are many others as well, but under this sub project it gives you an opportunity to you.

A

You know to talk to you, the IBM team members working in this open space in the Kuban YZ community right. You know many developers and leaders. Some of them are here. They regularly attend the Sigma things or disturb project meetings. We have some discussions on the slack channel. We provide updates on the /mo for this provider. Ibm cloud, it's on one of the Cuban ease slack channel regularly on the public side on the private side.

A

Also, you can follow the evolution right off the IBM cloud platforms with respect to kubernetes and some of the other cnc projects I mean you know: I've been clouds your leveraging, so many projects which are part of the CN CF. You know helm and at CDE- and you know many more right from it. Yes- and you know so- you know this this- this this platform provides an opportunity to interact with the team that builds and operates IBM cloud, alright. So how how it's, how it is created?

A

What's the structure of this sub group right as a project we have three co-leads I know it's like a fault-tolerant system here as well, and the odd numbers are good three kool-aid. So at least one is around all the time we have Kali dammit, who, unfortunately, could not make it to the conference because of the conflict, but he is specializing. The thing called MCM multi-cloud manager. We have resources, he he represents the IKS, the IBM cloud chemistries and rocks the OpenShift site on IBM cloud qumaris service and my subsidies Allah.

A

Our guru presents from the open source side. You can learn more about about the sub project in you know in more depth and go through the scope, the governance and what not I would just brought a link. But it's again the get up. Calm cuban. It is it's a it's coming to repo, so it's still called seek IBM cloud there, all others, you know the previous provider. Sig's are still called sig.

A

But if you check out this link, you will see that you know a big deprecated sign there. So eventually the documentation out there for individual providers will move under the cloud provider documentation. We have a mailing list which is called now cuban squadra bimcloud, like we have made updates as possible on different areas. Everything is updated with this. You know the change in the structure right, sig, verses, sub-project, so that has been done and then you can find more about this project again.

A

On the you know, we have good read me out there on the github community PO.

A

What do we do just highly? Well the activities we there are several ways you can communicate with the with this sub project of cloud provider. Sig you know and and participate right. We would love to have more people involved. You know more people contribute to the different parts of the project. So you know I mean, like I said we would love to have you contribute and would love to help you out.

A

You know if you are new contributors, you know one of the thing I do as part of my job is besides you know co-chairing days and other things and gets the maintainer one of the other thing. I. Do is participate in the contributor experience special interest group, and you know we always help new contributors getting on board. So you also like seek meet and greet session on Monday I, don't know if any of you were there, but you know it's basically having new contributors on board, and you know this project.

A

Sig's are one of the you know easier way to get on board, get familiar with the community work and you know that kind of stuff. So, if you're interested, please, you know reach out afterwards. You know on slack or you know here right, and you know we can talk more on that, but we meet every other week on Wednesday at 2:00 p.m. Eastern Time.

A

All the meetings are recorded in a hyperlink. They are all on the Cuban ease YouTube play channel there. So you can watch all the previous recordings there to learn more about. You know the past discussions.

A

We have presentation typically on various topics. You know from both public and private cloud side. We talk about like what's the upgrade strategy from from the IBM right Cuban, it is we talk about. You know some use case with with the users like sea snake. We have a you know, couple of folks from cystic. They typically join.

A

The call you know like nab folks and you know, I was told by folks that they usually like to watch videos later on, so they watch the videos slack discussions like I said this is slack channel called cig, IBM cloud or provider IBM cloud, it's out there and other thing we do is we we provide updates of our activities. So every other week we have a cloud provider, Sigma dings Y, that's the bigger one as I mentioned earlier. So all all this working group they represent their.

A

You know in from, like I said you know from AWS and from in GCP and Ezreal and OpenStack, and so all these traps are there. We always attend there before the updates. There.

A

We have sub projects under this working group. The first one is called cluster API provided IBM cloud. This is about the cluster API. Probably it basically leverages the cluster API project, which is on the github calm, it's again community project, but all other providers. They extend it. So we have, you know the work going on there again. If I interested in contributing place, you know take a look. Brad will be covering that more.

A

We have IBM cloud provider work going on. You know, that's the IBM cloud provider that we used at ICAST, the CUNY Service more on a load. Balancer and Richard will talk more. You walk you through about more interesting stuff. You know the architecture, the cloud controller managers stuff and you know other things.

A

One of the thing which the cloud Porter stick has been focusing for some time is moving the entry. So some of this border has entry code, which is part of the cuban is report they moon get out so color of tree to make it more. You know clean, so that also reachable talk a bit more on that and then organizer have mentioned. You know we participant the cig cloud provider general activities, we we participate, the cube corn and in the community meetings and whatnot.

A

So with that, you know with the basic overview, we're gonna talk a little bit on the or you right like I mentioned earlier, which were you of IBM cloud on the cloud companies stories, we're not open shift on IBM cloud. You know what exciting things happening there. We have multi cloud manager, so we're rich or will first walk you through about IKS and rocks, and then we have yawn I will walk you through multiple manager and you know, will continue continue talking and your discussion.

A

I have any questions, please, you know feel free to enter us with Richard.

A

Thank you. Thank.

B

You everyone so yeah I'll, be talking about IKS or IBM cloud community service and then what we call internally rocks, which is our Red Hat OpenShift on IBM cloud.

B

So I'll walk you through the services themselves at a high level and then I want to go into a little more detail on actual cloud providers and what they provide under these services and the cloud provider itself does not have to run as part of a managed service. It could be kubernetes on IBM cloud, but we'll start with IP assets are, if you want to say, vanilla. Kubernetes services are managed offering where you can create your own kubernetes clusters. Your compute hosts, deploy, manage your apps.

B

We participate in the CN CF kubernetes certification program, so we've been doing that since its inception and we started IKS in the 1.5 kubernetes release timeframe, and today we, our latest version, is 116. So we've come a long way in the certification all along the way from from I believe it started in 1/7 timeframe for the certification program to today on 116. So, as you would expect, we deliver kubernetes as kubernetes, since all the features and functions that you would expect to have from kubernetes are available.

B

Obviously there are certain things that are not on our IKS service. We don't turn on alpha features from kubernetes. You get beta and GA features, but alpha features in general are not turned on, but otherwise it's certified and the certification is that the GA level of api is that's. How coop does the certification only GAAP eyes get certified through that program? So if you want more information on the service, what it all provides, what you get with it I have a link there.

B

We can go check that out next thing, which is a pretty important topic. If you're using kubernetes is currency, keeping up with releases and so on now it's a it. I should have an asterisks on this because Kubb hasn't released it 1.17 yet so it hasn't had its fourth release of the year, but it will be coming here at December. 9Th I think is their target.

B

So anyway, I TS kubernetes we've had four releases in 2019, assuming we get to that 117 released from Kubb this year, which we should and one of the questions you have to ask yourself. Are you staying current, and this is a topic that goes well beyond IBM cloud kubernetes service? Any cloud provider has similar discussions.

B

There was one yesterday from Google talking about update strategies. It was a really good talk if you haven't had a chance, I'd encourage you to go check that out, but it comes down to the question is: does kubernetes need a long term support strategy for its releases and I laid out here? It might be a little bit hard to see, but if you download the slides a little bit easier at the bottom of the slide is IBM kubernetes services versions when they were released and when they will become unsupported.

B

So starting at the bottom of the slide, is 1.12 release. It came out in November of last year and it was unsupported as of November of this year, so November 3rd and then next is 113. It's marked as deprecated right now it will be unsupported in February of 2020 and then going up from there. 14, 15 and 16 are all supported, and the timeline for each of those is along that same line one year.

B

Now, if you look at kubernetes itself, kubernetes supports 14, 15 and 16 as their release they're under development right now in 17, and when 17 comes out, they may give you some patches for 1.14 release for a very short period of time, but then it'll be unsupported. So effectively, you have nine months of support once a release becomes available from kubernetes and for most folks, that's just way too fast. So that's where this question comes from a lot of community members is what to do next. Well, I've included a link here.

B

This is a recent update to the community on the working group that they have established for the LTS to say what does kubernetes look like in the future, with respect to support, so you'll find that, whether you look at IBM cloud or you look at Google or others you'll have a similar structure and release support from these other cloud providers as well, because there's not many folks that want to maintain multiple releases that kubernetes is not going to maintain themselves.

B

So generally, you have this same model where you get the latest releases, but you'll have some deprecated releases and then eventually unsupported. So with that the community is considering supporting four releases of kubernetes at a given time. So that would give you a one-year support window which would potentially extend you know our support window then, instead of having 1:13 be deprecated, we've potentially could support that much longer.

B

So a lot of good questions there and then on on that you'll find that there's some building blocks the community needs to establish before they could consider even a lot long term support. So if you have have interests as a user operator of your cluster, please watch that that discussion and contribute, if you believe you could help out in those discussions. So then let's jump over to Red Hat OpenShift.

B

So we just came out with it in August, Red, Hat, open, chiffon, IBM cloud, it's a managed service as well similar in structure to our kubernetes service. We manage the control planes or the master so that be your API scheduler controller manager and cloud provider at that level, and then you, you know pick your worker flavors and we deploy some critical components. You get what openshift gives you and then from there you deploy your apps. So you take a step forward on open shift from kubernetes.

B

You get all kinds of enhancements from that level to help you modernize and deploy your applications and manage them so a lot of tooling from Red Hat. To do that, we have a lot of details on our Doc's. If you want more information and all that all works right there in the link in the slide, we currently support the 3.11 release, RedHat OpenShift and it is based on Kubb 111. So there's little asterisks they're like kubernetes, doesn't support 111 anymore, but Red Hat does okay and it's their version, updated with patches security fixes.

B

So that's all still happening so, even though it is based on coop 111. Don't let that scare you it's getting all the security and update fixes that you need. We don't have a tentative on a supported date. Yet this is our initial release all right. So that's the manage service side of things, though I want to jump over to the cloud provider. Side of things and I'll call this our IBM cloud provider, KCM architecture- and it's actually applies to any cloud provider, and this is from the kubernetes documentation.

B

So you can go out and check that out there and it walk you through in a little more detail if you want to come back and refer to it later on what this all means. But this is a kubernetes cluster architecture on top of a cloud without the cloud controller manager, and if you look at the various components at the top left is the coop controller manager and it has a little arrow there.

B

That is a cloud connector, we'll call it and then API server is next and at the bottom right are the couplets talking to the cloud. So all these have connection points into the cloud. That's how coop worked from its early days and that's where was mentioned- that kubernetes has all this code from cloud providers in core kubernetes and that's not ideal. Obviously, if you look I beams, not there and the reason iBM is not.

B

There is because when we started our journey on the kubernetes service, they were not accepting new cloud providers into the codebase because they knew that the explosion that was going to happen. They didn't want to maintain everybody's cloud in kubernetes core, so that was going to be a problem, so we needed to change this architecture, but getting there is not easy.

B

So we need to disconnect these connectors in each of these core components of kubernetes, so that the cloud connections can be done in a more pluggable fashion, and so that's what our journey is a out in the cloud provider and a lot of the discussions are about within our sig and especially the sig cloud provider umbrella, as how do we go from here to here? This is the endgame that we want to achieve. So the endgame is each of those core components of kubernetes the controller, the API server, the couplet.

B

They no longer have cloud connectors. They don't want to talk to. Nor need to talk to the cloud. All they need to know is that there is a cloud provider, it's external somewhere and it will handle the core applications that need to happen on the cloud and that external cloud provider is the cloud controller manager. So the community wants to provide a reference implementation for this, from which cloud providers can build their own cloud controller manager and, as you would expect, there's Google and Amazon IBM others doing this.

B

You can even we had a cig, a deep dive yesterday and someone was asking about billing a cloud provider for their private cloud right. So you you there's nothing. That would say you couldn't do this yourself for your cloud.

B

The goal is to get to that end, game and and for most folks that start fresh, it's easier. If you're in the the area of extracting and migrating, it's gonna be a little bit more difficult. So what are the? Let's go a little bit deeper now into the interfaces. So what is a cloud provider? What is a cloud controller manager do under either architecture, there's a common interface that you implement and you can implement various aspects of it. You don't have to implement it at all.

B

I'm gonna highlight what IBM implements in our cloud provider what it provides on each of those interfaces and what it doesn't need to do and why it doesn't need to do it so at the at first is one of the biggest things cloud provider gives.

B

You is a load balancer, so each cloud has their own cloud load balancer and the integration point is through the service controller and kubernetes to say: if I create a load, balancer service kubernetes will set up a lot of the components like the cluster IP, the node ports, all of that routing in with internally with ruku proxy but ultim. You need a load, balancer created and that's done on behalf of the cloud provider. So we have an IBM cloud. We have three versions of our load balancer service.

B

In the early days on our classic infrastructure, we had a network load balancer that ran within the cluster. It was based on IP tables, with virtual IP managed within the cluster. Through keepalive d, we've extended that to an l network load, balancer version two which is in beta, is our IP vs base load balancer.

B

It also runs fork on the classic infrastructure and then, lastly, on our V PC infrastructure, we have a new V, PC layer, seven load balancer that we've recently integrated into our cloud provider, so load balancers are critical for most applications, as is obviously nodes. Instances is what the interface is in cribben at ease, so Kubb needs to know what been on what information about the nodes such as hey? Where is this snow and what zones it in?

B

What's it's IP, you know different aspects of the nodes, so kubernetes can do things that are really critical, like hey scheduling and cross zones, for example, and that's where also the zone interface comes in, so they we rely through the cloud provider on our bootstrap process to set up the node from which the cloud provider can extract the data and give that to kubernetes to finish the bootstrap process of a node, and now we recently enhanced our cloud providers.

B

So obviously, so we can get from the KCM to the CCM architecture through those two interfaces and then a couple more, not a lot there, because we don't implement them, but other other providers do so clusters and routes routes. In particular, we don't implement because we use calico in our managed services today for routing on the overlay over the infrastructure.

B

So we don't implement that other cloud providers do but, as I said, the each cloud provider can choose which interfaces they want to implement, and you really I mean you don't even have to implement the load balancer interface, if you, if it's not applicable to your environment, all right. So next steps migrating X. So we're because we're not in core base. We don't have to extract anything. Other cloud providers do, but we do need to migrate.

B

Now, if you're running rocks or the Red Hat up wind shift clusters, they already have migrated and they're using the CCM today. So no action required. That's already ready to go for our IPS clusters. We need to do the migration from KCM to CCM and we're hoping to do that soon. Time frame and 117 is hopefully, if we can achieve that. That's when we'd like to do the migration. There are some challenges with that. If you're interested in our status, we keep it posted here on this spreadsheet.

B

This is not just IVM cloud status, it's all all Klaw providers. They have updated the tracking sheet to say where they're at on their migrations to the new architecture. So some folks, some cloud providers have a lot more work to do. There were a lot more ingrained in kubernetes, so the challenges are a little more significant, but one of the biggest challenges for anybody. That's making this migration and our goal is to make it seamless like most folks that use any of these services shouldn't care that hey. It's got. This KCM cloud provider vs.

B

CCM but is locking locking, is one of the key issues for h.a, so kubernetes runs multiple controllers that manage certain resources. Service controller, for example, manages load balancers, and you only want one service controller running at a given time and they do that through leader election locking and at CD, and we want to make sure that when you migrate from one to the other, that you don't get a service controller on the KCM and the CCM running in your hea environment, trying to create or delete load balancers at the same time.

B

So that would be a problem. So the community is working on a solution, for this did to maintain availability, yet control the locking so that only one controller in each of the architectures is running at a given time, so that you can ensure that migration is smooth. The alternative is a little bit of a downtime, taking the controllers down to not have a leader on the service controller, bringing it up on the new architecture and so on. So that's an that's a big area.

B

The community is interested in and achieving to make this final migration happen.

B

So what's the future hold for the IBM cloud provider work? Well we're going to continue on this migration strategy to the CCM working with the community, where we need to to help that happen, and the community has been really good about it because they realize that there's a lot of folks already out there having a lot of clusters. They need to do this right and so they've taken very good care to help ensure that all cloud providers are able to move forward on this path together and not leave anybody behind.

B

So that's been really helpful for us and, like I, said right now, our cloud provider is just used in IKS and rocks and we love to hear if folks have interested in building their own kubernetes clusters on IBM cloud, and in that case, would you be interested in the IBM cloud provider and using that, of course, that would mean we'd need to open-source that. So that's one area that we're looking at with that. If we do open source, we need to do the improve documentation for usability in the next area.

B

We're looking at is- and this is a community initiative as well- is integrating cloud provider testing with the overall CI CD testing for kubernetes, so Kubb comes out with, say: 117 they're working on a release candidate one and they want to know what. What is this release candidate look like if I run it on Google or Amazon IBM? Whoever right is it still good quality or we've seen some significant failure rates across the cloud providers? So then you might have to step back and say what what changed here. So that's that's.

B

An area of ongoing concern in an area for improvement, is and we'd like to not only just submit our results through the certification program, but also into the general test framework of kubernetes that we can say, hey look, 117, looks good for IBM cloud right or this patch released 116 5 when it comes out, looks good too, and then lastly is container image. Security is another important topic.

B

If you look back at the various components, kubernetes has provided as part of the release artifacts.

B

Some of the images have a lot of CVEs and that's based upon the distributions are based on top of additional add-ons that are installed, and so we're always looking to streamline our container images that we provide through whether it be our CCM ourselves for the cloud provider or things that are provided by the community, say hypercube, which they are now moving away from to API group controller manager, scheduler images as well. So that's an important topic all right with that.

B

That should cover the cloud provider and I can take any questions now or we can do that later as well and I'll turn it over. Then to you. Thank you. Oh sure,.

B

So the question is what KCM. So the question was, what is kc m kc m? That's the coop controller manager, cube controller manager and I use it in a kind of a loose term on those slides back there IBM cloud provider, kc m. I call kc m architecture, which is really to say that the cloud provider is built into the kc m itself to provide controller services through that. So that's what that is, and then the CCM is the cloud controller manager yep anything else.

B

B

I would say that, well what would ask so? The question was a speaker yesterday would say: kubernetes is not highly available. I'm curious to know what aspect that they were referring to. It was a control, plane, data, plane, applications or so on. Right, there's a lot of aspects to hae there.

B

Okay, so more focusing more on the node AJ or okay infrastructure layers, so I think I think one of the challenges you got to think of kubernetes is: it's got a really nice architecture in the sense that they have state that they store in at CD, and they want to achieve that state and then controllers keep working to achieve the state right. The idea is that things will fail and that you need to be able to recover from those failures.

B

So if you use the tools that kubernetes provides you can you can help mitigate failure and thus achieve the h.a you want if you use the tools right. So there is a challenge involved to using them right.

B

So if I have a deployment 1:1 replica on one node and that node goes down yes, AJ is not good, but if you can say, I want 3, replicas I want to spread them across zones, and if one zone goes down then my application is still serving in the other zones, and you put that behind a load balancer and using kubernetes probes on pods liveness readiness. You can start to achieve the H. A you need to so I'll say from IBM's perspective, using the tools that ku has.

B

If you use them right, you can get very good H a through through kubernetes yeah.

B

Yes, so that question was a the the speaker mentioned. Sorry, I should get you the mic here, the that you can get security, but not for free, basically yep so yeah totally and that's that's it UK. It's the tools are there, you need to be use them and if you use them right, you can make it work. Yeah.

B

One more yeah might it might as well yeah that might help.

B

Maybe they'd like to hear it on the on the video that then then they know.

C

Just a quick question on rocks now that IBM has acquired Red Hat like how closely do you work with like Red Hat's teams on this? This is more like an arm's length sort of partnership just like if I was doing open shift on AWS or is it more like closely integrated like? What's that, look like sure.

B

So well, if you look at the time when we released rocks, which is August first and you think back, you know, we did development before that. That was before the the whole thing can really close and all that so we couldn't, we couldn't work together. You know closely on that future roadmap. So the question is, you know what is the future hold working together? Certainly, yes, I mean we want to. We want to work with red hat to.

D

B

D

In many ways view red hat and IBM is a partnership. Okay, because it's very important IBM that right, head, open shift, runs just fine on all the platforms yep, so I just want to be very clear: do we work with them? We work with them just like they would work with any other partner. But if you go talk to Red Hat, there's there's a lot of autonomy. There there's a lot of freedom there and and I think the best way to view it is as a partnership.

D

We, we obviously want openshift to run as best we can and IBM cloud and we're gonna ask some questions like any other cloud provider would but I do think it's worth folks understanding that when he goes through what Red Hat does their stuffs gonna run on all the clouds just like it always did right.

B

Exactly so, that's that's how we approach it today, right.

E

So follow up on that I wanted to ask red hat guy talked to me yesterday. He said, of course, iBM is saying you should use OpenShift and I said. What's OpenShift they've never discussed it with me. So I guess that's a totally separate offering that we can choose that. We just haven't chosen to look at so yeah. We just wonder: what's the urgency? Is that something we should move to or can B be perfectly happy? Never knowing about it.

B

B

D

The really nice thing about open shift is is a couple things. It's really focused on easing the pain in certain areas. One of the areas where openshift does a really great job is what it calls source to image. So it can pull content from the github repositories and build the images for you, take care of all that pain and the the animal deployments. So so a lot of that painful piece that, if you're not really experienced in building those docker files and UML files, it's gonna take care of that.

D

The other nice thing that OpenShift is gonna. Do is it's a little bit opinionated? It's gonna make some decisions for you on how the networking should be and provide configurations for the networking for you out of the box so, and it's gonna also provide you some good opinions on how to do your your continuous integration, continuous deployment, so the best way to think of openshift is its kubernetes underneath, but then making some opinionated choices and decisions about how to do some of the pieces that you may not want to deal with.

D

If you just want to go develop your app. That's the best way to look at openshift yeah.

B

Exactly and so it kind of it comes down to- where are you at in that journey? Right, I mean? Do you want that help, or are you like? Oh no I need to be the latest and greatest kubernetes I need that feature and I want to build my own right. You know, that's yeah, sure anything else.

B

All right, I turn it back over for real this time. Yeah really all.

F

Right away, Thanks, so I'm gonna talk to you just a little bit now switch topics, just a hair and just gonna do something quick, there's only it's not a huge crowd today, but I do want to kind of see what the outlook for your companies are. As far as number of clouds that you will deploy to or a number of clusters that you might have so hands up in the crowd, who has at least one cluster in their business today, like that, should be everybody right, yeah how about two yeah?

F

How about multiple providers so meaning local or on the cloud? Okay, all right. So at least one that's good, all right! So what I'm going to talk to you today about is a new product that IBM has announced. It was out the start of this year. This in I think was September. It was released on top of OpenShift. That is new the new platform of choice. It runs on IBM cloud. It also runs privately. The product itself, though, is meant to help.

F

You manage multiple clusters, so the idea that you could have in your environments, many different types of clusters and used for different things- is kind of the key to this product here, and so you may ask yourself: well: why would I want to run multiple clusters and can care about them and manage them? And so some of the things to consider when you start to develop more and more of your business applications on kubernetes? Is you know you may have laws you have to apply to for where your data may be right?

F

So some of that may be like I have to have my data within the country, or maybe it has to be on premise or I have to have it at least in Geo's, like I know, certain banks and stuff have to have certain rules where data is replicated across to different geographies. So so those are things where you may consider. I may need to have a separate cluster in different locations right and then manage the information back and forth. The second piece is: okay. Now I've got my sensitive data covered.

F

What about my very expensive large amounts of data that is not private, right and I'm spending a huge amount of money running that locally in my environment you know: buying storage, racks and whatever I. Don't need to do that. So the answer is, is you know you can run that some of that in cloud services and those cloud services will generally be cheaper than local we're not going to be opinionated on which one you use. You can negotiate your own deals.

F

Ibm would like to sell you that, but you know there are other companies out there. That'll do that, and the point is MCM will help you kind of manage all those clusters across that environment. So a fourth or through story. Third reason would be services. So, as you start to build your applications, you may be looking to say use some analytics, say Watson analytics or maybe you're after goo, google's AI or some object, storage and amazon. Just because you know your app developers are gonna be out there.

F

Building cool stuff and they're gonna want to use the latest and greatest stuff, and if you tell them it's in one cloud, they're gonna get really angry and get frustrated right, so they're gonna want to be able to run stuff where they need to run it and and with the services that are available today and of course, the last one is kind of a disaster, recovery, resiliency kind of story and that kind of plays well into you know you may not be in more than one provider, but perhaps in different regions.

F

So you could have you know within LA and then maybe another one in Texas, so that if tornado hits or you got an earthquake, you don't have a problem with data loss. Okay, and so those are some of the reasons why you would want to have clusters spread across geographies or maybe even just you know- maybe not geographies, maybe they're close together, but they're offering different services. Okay.

F

So what does multi cloud manager do? And you got that case, so it does a couple of things. So, first off you can see in that little picture. There there's kind of an overview of your cluster information. You can kind of start to see where you have capacity.

F

You can see where your compliance is set to, and in order to do that, of course you have to configure policies, you can say, and so in the bottom, in the kind of the middle there in the blacks, I, don't know if you can read it very well, but that's an example of a yama where you can configure a policy where you can set the container size to not grow past a certain memory allocation right and then, of course, there's like a little diagram here right next to that one in the bottom corner.

F

That is a part of the app model, so inside of MCM they're attempting to pull together what they call an app model and I have another slide on this in a second, so I can get really deep into it. But the idea of is it to pull together your clusters and your your helm, sharts and all the other pieces into some sort of a view that helps you represent the actual application that you're managing right and so one of the things that will end up happening over time. Is you may have a cluster?

F

That's local running your data and your database access, and perhaps it has a you know: WebSphere MQ or a Liberty server running to host Bandol back-end processes, and then you have perhaps out in the out in the wild while west of the cloud you have. You know app servers and other things and load, balancers and stuff like that running your web mobile business and you have data transferring between the two and so in the diagram here. I, don't know if you can see it there's kind of in this example.

F

They have this application running across four different clusters and two of those being cloud. Ok, so so those are kind of some of the tools that help you kind of manage that environment as it gets more and more complex because you're gonna get past. That point. Where it's hey, I'm, just gonna log into that cluster and see what's going on, I mean you know. If your application team comes to you and says, I have a problem: you're not gonna, know where to go right.

F

You've run in five different clusters and what's going on, is a you know, a node problem. Is there an application problem? Do you have networking issues? Those are all problems that you need to have some sort of a tool to help. You resolve all right, and so this last chart here I think this is where Khalid was asking for your help. So what what's going on is we're building this application model that I described earlier, so this you guys are probably familiar with helmet arts right.

F

So if there's a helmet chart there and we've added an app yeah and an operator Gamal as a part of it now what happens is inside the app Gamal, it kind of configures all the pieces and the policies that are a part of it and then the second thing that it does is it sets up subscriptions. So this is a new idea that we've kind of pushed and it's now open source, so we're asking people to contribute, and so the links are below here. So there's an e operator subscription and the operators deployables o.

F

What happens is when you deploy your application, it registers with the subscription and this there's a channel that is constantly watching for changes in those subscription pieces. So, for example, you may be, you have a subscription to a load balancer, a certain type of load. Balancer, that's running and your subscription may say you know with you know, point releases I want updates right, and so what should happen? Is you have a periodic check on your at your pod level? It will continually run using the operator model and checking to see if there's updates.

F

Now, when there's updates running, if you have in your app model to decide, you want automatic updates, it would do a rolling update using the operator path and actually update your system, keeping it live and running and getting the latest frequency updates. Now, that's again subscription-based, so you can control whether it happens or not. You can also control whether it's automatic or it lets you know in the MCM that there's a you know, it's something to be done and then you can trigger it to change.

F

Okay, are there any questions about that or anything about MCM in general, I can kind of take because that's kind of the the last of those slides.

F

Okay, thanks for you time.

D

Okay, so um one of the things we look at is a hybrid multi-cloud and looking at a strategy of kubernetes running everywhere or OpenShift running everywhere. So look at how we're gonna support that there is a kubernetes sig for clusters and a cluster api and the different providers need to provide their sort of cluster api provider. It's an interesting model that the cluster API, if you're not familiar with it, it's open source. But basically what happens? Is you have little bootstrap cluster and from that bootstrap cluster?

D

You use that to actually get your actual real target cluster up and running wherever here it may be, and in our case it's on IBM cloud, and they have you know just like you have coop control they've got cluster control, so they've got commands and there's different providers. All the different providers provide their implementation of the cluster.

D

Ibm provides it and we have several folks who are contributing to the open source project and in kubernetes for the cluster API, so that work continues to go on and we continue to focus on it from multi cloud right, so multi cloud and if you're able to do it through the cluster API. That's that's that's the way to do it.

D

I want to go into too much detail there, but it's it's sort of a piece of community that if you want to take a look and see the work that we're doing there, gwon Yalu is one of our leads. Who makes a large number of contributions to this project and several folks on his team do as well. So just a little piece of what's going on there for hybrid cloud and cluster, and now I'm gonna change gears a little bit and talk about some other open source projects that that IBM cloud is really interested in.

D

So kubernetes is great cloud native computing is great, but if you look out there it can get a little complicated. Now, that's what this chart. So if you want to do cloud native right, you really need to understand the concepts of 12 factor. Applications understand micro-services, understand all the different projects that are available in the cloud native computing foundation, there's different choices, different options, and some of us really enjoy this right. This is our world and we live in it and we really enjoy it.

D

But again, you know put yourself in the shoes of a Java or j2ee developer, who wants to get all the benefits of cloud native, but maybe they just really want to write their Java applications and really maybe data, maybe they're. You know it's nothing wrong, maybe they're, just not quite as interested in all the nuances and details of cloud native like some of us are, it could happen right.

D

So you know the goals and the challenges we have is: how do we make those people happy, because every java application developers they're out in the world a lot? So what are we doing? We've put together some open source projects that are really tailored towards helping to bring the benefits of cloud native to folks who work and say Java or Python, or Quercus or node different different areas where it's pretty straightforward.

D

The person really wants to worry about their code and and get the benefits of cloud native, and how can they do that as painless as possible? And so one of the projects we have is habanero and it's got a lot of associated projects with it we'll kind of talk about what they do.

D

But it's it's gonna provide some cloud native frameworks, integrated development tools, automated DevOps tools give you some options to help you modernize your applications and move them to a cloud native environment, and this is all built on standard best practices for for running in cloud native environments, and so one of the ones we really like to talk about is AB City. So what AB City does is out of the box.

D

It's gonna provide you different runtimes it'll, provide you a WebSphere, Liberty runtime it'll, provide you a Python, runtime it'll provide you a node, runtime it'll, provide you Korkis, runtime and so it'll provide you all these different runtime stacks, so that in your enterprise you don't have everybody building their own java container stack right. You know.

D

Hopefully you can get agreement, consensus and there's you know your architects who can decide hey if you're gonna use Python here's the runtime stack, you should use if you're gonna run java, here's the stack you should use if you're gonna use node, here's the stack so at least some way to get some sanity and consistency across your enterprise is the idea and so App City will probably it's it's. It's it's built on operators.

D

So it's using best practices, for you know lifecycle management in in kubernetes and with this people can really easily get up and running on on those stacks and other parts of what was on just the previous chart. We've got integrated development tools, there's a project called code winds and it has plug-ins for both Eclipse and visual studio code. So in your favorite development environments, you've got the ability to start getting some cloud native help to to to build those applications. Get them turned into images and deployed out on your favorite infrastructure, kubernetes openshift what-have-you.

D

The other thing we've been focused on IBM at IBM for the past 18 months is getting all of our middleware to be cloud native. So we have taken responsibility for saying.

D

If you look at all the enterprise customers out there, they're using lots of enterprise, software or db2 MQ, WebSphere Liberty, for example- and if you look at what we approximately is about 20 percent of that software out, there has been able to make it to cloud native, but there's that 80% kind of lagging behind there's a lot there, and so our goal over the past 18 months was to get all of our software cloud native and and ready to run on the cloud. And so we've done this and we call these cloud packs.

D

They're, IBM certified, run anywhere. They're gonna run on all the different clouds run on OpenShift and OpenShift runs on all the clouds. So you can feel very comfortable that we're getting all of our middleware so that it runs anywhere and that's why we don't necessarily care where you run it, because we're gonna make sure it runs everywhere and run on top of OpenShift.

D

And so we've got several examples of these. The cloud pack for applications, the the big piece that you'll find in there is the WebSphere Liberty.

D

So if you're wanting to build j2ee applications, you know Java all that's in the cloud pack for applications cloud pack for data, you can think more like your db2 and other databases and database related type stuff cloud pack for integrations can be more like your mq family of middleware, and then we've got a cloud pack for multi cloud management, so we're also making sure we've got pieces to worry about running multi cloud, and you know people say well why multi cloud right well, we've seen these stories before think insurance.

D

Companies you've got one insurance company, it's use in one cloud: it's got a lot invested in it and what happens? Another insurance company buys that insurance company, and maybe they were using a different cloud so, whether they like it or not, sometimes because of how the world works, you could end up using multiple clouds, and so you know we worry about providing offerings for our enterprise customers, so they can deal with those types of situations.

D

Sometimes you may have customers that want to use multiple clouds, just like in the old days where they would sometimes have an or half half Oracle databases, and some db2 database is because the enterprise could then have a little bit of leverage. Hey I could go either way. I could use this database. I could use that database same thing with clouds, so you know we have an opportunity to see multi-cloud opportunities and and we're investing is there in there as well.

D

So at this point, we've covered a lot of stuff and so I'm going to open this up for discussion and any questions. Anything we've got several folks, smarter than me that that we can force them to answer those questions.

D

Okay, well, then, you you get to go.

D

That's a great question, so would you repeat the question: what's that yeah I'd want to give him a? Do you want me to repeat it? So the question really was about Cloud Foundry and how does that fit into this world? We haven't really mentioned it.

D

Cloud Foundry is something that we offer and the IBM Cloud there's a Cloud Foundry service. We've done a lot of really cool things with it, and you know what we've been looking at in. What we offer is how we can provide cloud environments that allow you to use all your Cloud Foundry stuff when you want to use it and how you can also use the non Cloud Foundry stuff.

D

If you build some new stuff on kubernetes, so there's actually a project in a cig group in kubernetes called Service Catalog and what SIG's Service Catalog does is it's got what's called a service broker, and so it's got the piece inside of it and Jonathan Burke on is the chair of it from IBM. But what it worries about is how do you pull in all those Cloud Foundry services into your new cloud native world? So you know the point of view from IBM is hey.

D

You know our clouds are going to support both Cloud Foundry and cloud native stuff. We don't care which way you go. There's any of it. You go we're gonna make money, because it's all utilization right. So we we, we do a very good job of supporting Service Catalog and you know the techniques are that are out there. Now, there's gonna, be some more advanced techniques. I think you'll see coming down the pipe with operators.

D

That will also worry about supporting Cloud Foundry services as well, but but there's definitely effort there to make sure that the Cloud Foundry stuff is certainly at least in IBM cloud is, is something that you can continue to use and and use alongside with everything else, and then the way to pulled into the kubernetes cloud native apps is through the service catalog. So you know you can use it the way it is and then you is it pulling it into your new stuff. Does that sort of help?

D

Okay, did you have any other questions.

D

D

Well, and and now do you use Webster, Liberty or just big WebSphere.

D

D

So the answer is that that's the reason we have those cloud packs is to enable that was her, so I, actually a few years ago, led a team that worried about both WebSphere and db2 and how we could containerize them. And you know early on.

D

There were a few problems and particularly for db2, we needed some kernel patches and what-have-you, but you know, ideally with with you know, if you want to look at using the cloud pack version that that's what it's there for and they'll have Webster and what's really birdie for you and and and that should be fine.

C

D

That's what we want you to keep doing so so our answer is yes and we're gonna provide a lot of stuff to help you do it, because that's that's what we want to do. We want to get all those applications running, because once you get it to the to the cloud native and running the containers now I can start talking about 12 factor.

D

Apps I, don't know if you remember 12 factor, apps a lot of great principles for how you scale things up how you you know if things crash they'll, restart all these great ways to write robust, very scalable, web or cloud applications, and once we get you that first step right, then we can take you to the next journey, which is, let me show you how you got multiple copies of this running. Let me show you how you can scale this up.

D

Let me show you how if some of them crash we'll start new ones up for you, so you'll end up getting a lot of benefits once you've been willing to take that leap of faith and get it to the containers. One of the best demos I gave a Riley software architects conference is, you know. We typically show that the benefits of running cloud native, where oh look, we can scale it up. We can crash a few things and I start with a a pure cloud native application.

D

That's all brand new and it was a you know. It had a little language bot to do some voice, stuff's all fancy stuff, but the really. What the audience wanted to hear is well what about my middleware and then in the middle of the demo.

D

I go well, let's go do this now with MQ, because we had MQ already containerized and now you got to see the same thing, but now as with MQ nodes up and running and then when they crash being restarted and and using all the it used like stateful sets and and what have you in kubernetes. So that's really. The beauty of this is it's not just for the new stuff that you get all the cloud native benefits.

D

You know work with the IBM folks with cloud packs for the different cloud packs and let's get the old stuff running in the new world as well.

D

D

hmm Okay, fantastic. Did you stop by the booth yesterday and get a book or the other day? You got a book right, fantastic, so say you asked enough questions. I had him on my backpack, you were getting a free book, but you already got one excellent.

D

Did you want to worry more about the questions he had or we covered them.

F

So I mean I. Honestly I was answering my team with a question question they had, but you were talking about multi cloud management and and worrying about running your application across two different clusters and all that kind of concerns that right, yeah, so I think it's actually best practice. I'll be honest with you, I mean if you were back in the data center days running your WebSphere servers. You know if you were a bank or something you would have your applications running in two different zones: I'm doing backups and stuff like that right.

F

That is what you would do exactly right, correct, yeah, yeah, correct, right and- and so you know by having you know, you don't need to have multiple providers because you can do local, local or you can do you know IBM cloud or Amazon Cloud zones right and so like regions within so like Amazon. uh Sorry like California and Austin right, which is one of the customs I work with. Does that?

F

And so one of the things to watch for is making sure you have good data pipes between you know: you're gonna pay for that in a cloud service, but that's very, very important. Cuz you're gonna want replication at high speeds, and things like that right.

F

The NCM product that we had I touched on earlier might be very useful for you. If you can build an app model on the top layer to see your apps across the different clusters that you, like your different Web Store clusters you have deployed.

F

It may help you kind of understand if you're within policy, even so like some of those policies can be very specific. Like nodes are not going about certain memory thresholds you're not having restarts you're. Not so like those policies can be like security. They can be how well your application is running, and things like that right.

F

So managing policies, if you're playing in kubernetes and you're, trying to do production very important, it'll help you stay compliant and I know: you're gonna have like PCI compliance and you're gonna have gdpr and all these other concerns right and those are valid concerns and those are things you can use. Policies to help you set as well as just the you know is my application running healthily right. So those are some of the other things I would say is I do start to do these, like multi cloud or even just multi-site situations.

F

Those are things that I would worry about it and, of course, like we offer a lot of tools to help with that and I can talk with you a lot more about it and there's even presentation later at our booth, where it might go over somewhere that as well.

D

F

A

No more questions yeah, we are done, I mean thank you so much. Everybody know we we utilized more than an hour of our 80 minute marathon. You know so thanks for sticking around lots.

F

Of visualizations.

C

A

Great interactions and we look forward to collaborate with you more, you know, like I, said, join us on the on the slack Channel we have for the for the sub-project that are, for your sake, Mehring list. You know what not right, just let's keep in touch and we'll talk more as we go further. Thank you so much everyone thank.

D