Cloud Native Computing Foundation KubeCon + CloudNativeCon North America 2018 (Seattle), 16 Dec 2018

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Deep Dive: IBMCloud SIG - Sahdev Zala & Richard Theis, IBM

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

Deep Dive: IBMCloud SIG - Sahdev Zala & Richard Theis, IBM

The presentation will cover an overview of the SIG-IBMCloud and highlights of the SIG activities. It will provide details of how the SIG and IBMCloud contribute to the upstream work in various areas like scalability, networking, cloud registry, Helm and SIG App. We will also provide updates on the cloud provider and near-term roadmap for the SIG. The presentation will provide in depth coverage of SIG-IBMCloud activities, walk through, user stories, SIG upstream contributions to the Kubernetes and related projects. The details are still being worked on and we will revise with more details later.

To learn more: https://sched.co/Gre1

A

Hello: everyone, my name, is Sudhir Jalla. Welcome to the sink IBM cloud dipped our session, thanks for coming I work at IBM as senior engineer and also contributed kubernetes and at CD and I'm, one of the co-chair for the sink and.

B

Hi I'm Richard theis also a co-chair for the Sagan I work on the IBM cloud, kubernetes service on the networking team and also on the kubernetes delivery and pipeline team.

A

Alright, so here's an hidden agenda. If you were here for the cig intro session, you know we talked about six structure and in a little bit deep in the overview, but today we're not gonna go more in detail, however I'll just pre-fit in case you haven't, made it to the intersession, then we're gonna talk about IBM cloud. Cuban histories we'll call a bit more depth into the cloud provider architecture and you know how we are adopting different releases of kubernetes in I.

A

Guess we'll talk a bit about IBM cloud private and we're gonna talk about the contributions that are made by the teams, the IBM cloud team. You know as a in general, we have so many contributors from IBM I've been shown the top contributors in companies, but we're not going through all those contributions. We will talk more about the contributions from from the cloud teams from the secret perspective and then I'm going to summarize and would like to have some discussions around.

A

You know how we can make it better, any comments from you and what you would like to see in the future, cig meeting sets etc.

A

All right. So it's a sig which provides an open platform to discuss things around Cuban IDs in both IKS and IBM cloud. Private ICP. As you know, you know the conference program, it's it's. It's a big thing in the community in the communities we've been talking about it in different meetings and in past three days at the conference, so both I K as an ICP. They participate in the conformance program and the both are certified and the Sigma ting. As you participate into it.

A

You know you got to get interact directly with the team that builds and operates IBM cloud, both ICP and IKEA site. You also, you know this is an opportunity for you to follow the relations that happens in IPS and I, see ICP platforms, six structure we meet every other week, every other Wednesday, all the videos for meetings are recorded and they are on the YouTube. So here is a link. If you like, you can just go see all the past discussions. There. We have three sig lids.

A

We tried to have it a diversified group of lid. There we have Richard. This was also a co-presenter today he's from IKS. We have Khalid. A Madhu is IBM distinguished engineer, he's inclusion I here today because of conflict, but he leads from the ICP side and myself from open-source side. We would like to have even more diversified leadership going forward.

A

You know having someone from from non-ibm aside from different company, so if you like, please do participate in the cig meetings discussions and you know if, for the leadership leadership eventually that ladder the Charter is out there, you can take a look, and you know, as I said, would like to have more people join the cig, mailing lists and you know have discussions in the SiC channels. You can read more about the cig at the link, provided this presentation is available online and the complex web page, so you can download it later.

A

This is just an overview of what we discussed in the cig meetings. Besides the discussions on the slack channels- and you know my list- it's typically around Cuban eighties, as mentioned earlier- we we don't have any commercial discussions on IKS or ICP in the meetings there are separate channel for those, but we talk about. You know how we are in Cuba DS. You know how was the upgrade strategy or things like scalability, you know, and then we also talked about the community side of the sea Krait, how the work going on like maintaining the sea.

A

We are working with the cloud for other sig or eventually, all other six, the cloud for six main movers and sub projects. We would like to do that as well and then any you know contributions whether it's a PR where there's issues and discussions that are made by the sig members into the community.

A

So that's just an overview and things you know at high level, like hey Richard, now for the IES, alright.

B

Thank you yep. So, if you're not familiar with IPS, it's our IBM cloud managed service for kubernetes, and our goal with that service is to provide the management for you to give you a good user experience. So you can deliver your applications on IBM cloud and then leverage IBM cloud services. You know like Watson, AI and weather and so on, and some of the new services that folks are now checking out, such as log DNA or sis take which are pretty cool, so they're available. The IKS service is available in six regions worldwide.

B

Twenty-Five data centers and it's kind of the foundation we take forward with you know as part of my world on the SE, is how are people using kubernetes there and what we can leverage from that and what we can learn and bring back to the community to iterate on that operating kubernetes at scale.

B

If you want more information, there's a link there, you can check that out and get our Doc's now one big important part of our of our managed services, certification, of course- and we want to be certified at all times, so we always submit results even at the patch level. Well, when we have updates to ensure that IKS is certified conformant and passes all the tests that are in the community, so you can see right now. We have three releases and support. I've laid them out here, one 10, 11 and 12.

B

Now, with that, a more of an interesting topic that I been dealing with as of late, is the kubernetes release cycle in how it pertains to people keeping current, because this is a really important topic, especially with recent security, vulnerability in kubernetes and the fact that that vulnerability was patched by the community back to release a 1.10, but not it later because of the support structure that kubernetes community has they, they support three patch releases and then working on one actively in development.

B

So it is an important topic because folks are are struggling at times to stay, current, we're seeing it in IKS and and I'm sure, other cloud providers and and other users are noticing this pain as well. If you just look at the year, 2018 here, I've laid out that on this chart the release structure or not to really structure that the release dates for kubernetes and then IKS and Delta for that and where we are an unsupported releases on IKS. So kubernetes has had four releases one 10 through 13 this year.

B

Iks has also had four Buller slight delta. There we're working on 113 right now and then, as far as unsupported is concerned, we have 178. 9 are now and supported for IKS. So those things move quick. So you can imagine something that was supported at the beginning of the year by the end of the year, you're in urina here- and I heard, if you're not keeping up so there's a with that. There's a community discussion going on about what does it mean to maybe extend support longer or have something as an LTS?

B

And this was an important topic to kubernetes itself and to certainly to the cloud providers as it is. It does impact them. You know directly and then obviously, as an end user, that cascade stone so I have my own opinions on the LTS idea and whether or not we should speed up slow down. I mean I kind of feel. A LTS just put passes the buck a little bit later in the whole process. To make give you an out sort of speak and then what happens when you really reach the end of the LTS?

B

What happens next so I mean I'm, certainly interested to hear back from our users community on this and I know kubernetes community as well. So if you're interested in this discussion feel free to talk to me now or after and and also to get involved in the community, because the cig release has a working group around this.

B

With that, so op not only is releasing kubernetes on the top of our priority list, but also taking what we learn while running kubernetes out of production and scale and getting some areas that maybe we need new features on such as a networking area and egress network policies, and so on I mean these are important things for our customers, so we take.

B

Those in is just some of the areas where we collaborate with the community on what what kind of features we need in IPs and what we kind of what what needs to go back to the community for support for everyone else, so networking, scalability storage. These are just some of the areas and we certainly contributed in a lot of other areas as well.

B

But I want to talk about in and all more detail. Now is the cloud provider since it's a pretty key component to when we deliver kubernetes on IBM cloud as part of the kubernetes service, the cloud provider is our interface between kubernetes and the cloud, and this is no different. This architecture diagram is a general diagram from the community.

B

It pictures the kubernetes controller manager architecture also known as the KCM, and you can see this is the current architecture used by most cloud providers where the connection between kubernetes and the cloud is done in various locations. So you primarily have the connections in the coop controller manager to provide implementations for some cloud features such as the load, balancer API server, as well and on the couplets side. There's connections in there to the cloud to allow you to manage nodes.

B

So you can you can look at this diagram and say: well, that's a lot of connections in it and we need to simplify that and that's certainly the path forward on what the community wants to do now: the cloud controller manager or the coop controller manager code. Here we build in our cloud provider into that code as well as the API server in the couplets and that's what a lot of the other cloud providers do as well. So what what's in this so beyond this would be in the architecture.

B

So what actually does IBM cloud kubernetes service do with these interfaces so I just just laid out at a high level on the the big bullets there are the interfaces on the cloud provider and then sub-bullets describe what we're doing at those interface levels. Okay. So the primary primary interface for us is the load balancer interface, which we implement our load balancer for IKS. It's an in cluster load, balancer service, leveraging, kubernetes resources.

B

So if you spin up a load balancer in IPS, it actually creates a deployment in kubernetes that manages a VIP using keepalive, deep and and it provides the routing in cluster using coup proxy.

B

So that's on the overlay on the underlay and the IBM cloud network, the that is also managed, that's managed outside and on the cloud networking side so that, along with we have a couple versions in the latest release 1.12 we have a beta version of an IP vs base load balancers our first version and as our default is an IP tablespace load balancer in cluster.

B

Beyond that, we we do basic implementation, stub implementations to complete the cloud provider, requirements for instances and node management and then zones to primarily have our IKS bootstrapping process handled the rest of that work and then there's a couple other interfaces clusters which we don't enter and implement and then routes. Now we use calico and IKS. So that's our overlay network, the underlay provided by the IBM cloud network, but so since we provide the overlay with calico, the routing and network master sets everything up.

B

So we don't have to do any additional implementation in the cloud provider to handle routes.

B

So that's what it is today. This is the architecture of tomorrow and where the community is heading not only for IKS and our IBM cloud provider, but also other cloud providers as well. So we want to transition I'm going to slide back a couple slides here.

B

This is the current way and you can see on the master side, the components there for the controller manager, API server and scheduler, and then, if we move forward to this architecture, you'll see the cloud controller manager comes into the picture here and that that's going to be responsible for connecting to the cloud and performing those same interface operations, but on more of a well-defined boundary so that you can now decouple the cloud controller manager and the cloud provider code from kubernetes. So there's a lot of benefits to this.

B

First of all, cloud providers don't have to build and carry each other's code around we get smaller binaries in the end result smaller container images for deploying these these components and a nice cleaner interface for everyone involved. So we kind of leveled the playing field for all the cloud providers, but it does.

B

It is going to be an important part of this whole thing, and not only just for IBM but for Google, and others is that we need to provide for folks that are on this architecture, a clean transition to migrate them without downtime to this architecture. So that is one of the bigger challenges as well as code restructuring, the interface managed management, and so that's gonna be one of the higher priorities for our sagen 4 for IBM this years. With respect to our cloud providers to achieve this goal,.

B

So I and I put a TVD here on what do we actually have to do from a load balancer perspective on IBM cloud provider? We really shouldn't have to do much there. I, don't anticipate changes, changes more likely will come in the management of nodes and zones, as we no longer run in the couplet. You run in the cloud controller manager and nothing in the clusters or routes. So that's a high-level any questions on that. Otherwise, I'll turn it back over to side. Ivan will head over to the IBM cloud. Private I.

A

Just want to mention here the open source, open standard sides of the IBM cloud, private that it's it's built with open standards. As you can see here, it's based on kubernetes. It runs the country docker. It uses a darker engine to run your contour eyes. Darker images, helm is something we use heavily. There are several hand charts out there on the Gita IBM repo, for you know a lot of IBM, middleware and other applications.

A

Terraform is the layer. It's another open source project, if you don't know, but that's what we use for posing at the infrastructure level.

A

Here is a simple architecture: diagram for the IBM cloud, private, as you can see right at the bottom that you, it runs the master on the x86 and power Linux and the third one which looks very incomplete, there's in Linux, which is you know right now. You you can that's where you can run your worker nodes and you can mix and match which in all this platform, but the support for Z Linux to run your master. It's it's just coming sooner.

A

I think it should be announced soon in in maybe by end of this month, and we please that you know as your as your cloud gets bigger right. The number of nodes you are running just one master node is not the right choice.

A

You can have multiple master nodes for H a you will be using a proxy node on its own and then management nodes right so that you do not load the master so much so typically, it has not just a master node, but perhaps nodes and management nodes, and then you've done worker nodes on top of it and as I mentioned earlier, all IBM product services out there for you to use- and you know you can run your app workloads on top of it.

A

This is just some quickly mentioned that it comes in three flavors. There's a community edition for you to use for free. You can just run it on your laptop if you're interested and play with it, and there is a cloud native editions where, as it's mentioned here, it's mainly for enterprise developers, it comes with micro, climate or micro service build earlier to to develop your applications and it comes with lives for liberty and other IBM softwares.

A

If you go on about Edition that which is Enterprise, one comes with, you know: extra software's there I want to mention here that the contributions, the discussions, IBM cloud, private team- is doing in the Cuban airspace. As you can see those couple diagrams in this laughter, those few months old but obvious what the top contributor has mentioned earlier, we're not going to talk on those things. We want to talk about again the contributions or engagements from the IBM cloud. Private team with the community.

A

So first, when there is equipment, is Federation with you as we know that as you move to using Cuba Nadi's more cloud native technologies right, you know, but I have just one cluster that'll be multiple Cubans cluster around and down the road you want to. You know do things like resource sharing right built between multiple clusters. So that's what the work is going on. We have three guys participating there. It spins ideas like number three there helm as I mentioned.

A

We do we use hem heavily, and you know we contribute back to the community class registry is another small project where which basically provides catalog for API is for your Titleist tournament. We have IBM multi cloud manager and that's what uses the cluster history so with that you know. I just want to summarize the talk and then just have some discussions. Any questions from you, but it's a sig for open discussions around Cuban, eighties on and I, be in public and private clouds, which are IBM cloud. Cubed is service and IBM cloud private.

A

We would love to have more folks joining the sig and you know, participate in the discussions. Help us get better as I mentioned earlier. We would like to have a you know: chic leadership from nan IBM folks as well. So if you are interested you know please join, and you know we can work on the leadership side.

A

Ibm cloud teams contributes a lot into the community and, as Richard mentioned earlier, around our near-term goal. It's under investigation right now is to open source lab IBM cloud provider, interface and architecture earlier he walked through. With that, any any questions, any anything you want to discuss.

A

A

Well, that's where we are that's an optional options for the customer right around the IBM cloud, private! Well, right now, I think it's! You know we that was a big news, that a data acquisition right right now, it's business as usual. Radio at is ready at IBM is IBM as far as I know, it's my personal instrument.

A

So what are you offering you I mentioned from IBM side? It's going to be what I just and creature explain we I'm not seeing anything. You know down the road or unless the integration happens and I don't know, what's going to happen in the future right now, it's business as usual.

A

Anyone to add on that.

A

Cyril and Europe you I mean from development side or in general development cells, development, side.

C

A

One second is.

C

It on here so does the the easy answer is thousands and it depends on how you slice and dice the word cloud, because we have the public cloud team and we also have the hybrid cloud team that works on the on-premises ICP deliverable so between the two and then it further it. It depends on if you want to include the developers that would be producing cloud services so, but even with cordia's cloud platform, team you're talking several thousand from development perspective,.

A

Any comments on IBM Burnett and even when I had out there I mean to meet.

D

No, so we love open source. Redhead loves open source, we're all gonna love open source together, yep.

A

Just making open source better yeah.

C

Yeah so fundamentally we had the same strategies right, and so it's a matter of being able to combine forces I'm personally, really excited to bring in a lot of the culture for your open source and you're in your contributions, and things like that, so that we can add that you're, your developer outreach programs are fantastic.

C

We both have fantastic engineering teams, we'll combine those together, reassign squads and then and then take over the hybrid cloud market, in my opinion, but those those details of that are being discussed. That is it independently. It's for, for legal reasons, they have to be discussed even independently at a at a company level.

A

Good good questions, any other questions.

A

Say join the cig, you know be the part, be part of the discussions. We would love to have you guys for the future meeting slack, but not all current discussions, so any future.

B

Yeah, thank you for coming. Staying late on the end of the conference here, yeah.

A

Yeah all right well, thank you so much everyone all right! Thank you.