Table of Contents
outline true

Overview

ICN BP family intends to address deployment of workloads in a large number of edges and also in public clouds using K8S as resource orchestrator in each site and ONAP-K8S as service level orchestrator (across sites).  ICN also intends to integrate infrastructure orchestration which is needed to bring up a site using bare-metal servers.  Infrastructure orchestration, which is the focus of this page, needs to ensure that the infrastructure software required on edge servers is installed on per-site basis, but controlled from a central dashboard.  Infrastructure orchestration is expected to do the following:

...

This document break downs the hardware requirements, software ingredient, Testing and benchmarking for the R2 and R3 release for and provides overall picture toward blue print effect in Edge use cases.

Goals

• Generic: Infrastructure Orchestration shall be as generic.  Even though this work is being done on behalf of one BP (MICN), infrastructure orchestration shall be common across all BPs in the ICN family.  Also, it shall be possible to use this component in other BPs outside of ICN family.
• Leverage open source projects:
  • Leverage cluster-API for infra-global-controller. Identify gaps and provide fixed and also provide UI/CLI for good user experience.
  • Leverage Ironic and metal3 for infra-local-controller to do bare-metal provisioning.  Identify any gaps to make it work with Cluster-API.
  • Leverage KuD in infra-local-controller to do Kubernetes installation. Identify any gaps and fix them.
• Figure out ways to use the bootstrap machine also as workload machine (Not in scope for Akraino-R2)
• Flexible and Extensible :  
  • Adding any new package in future shall be a  simple addition.
  • Interaction with workload orchestrator shall not be limited to K8S. Shall be able to talk to any workload orchestrator.
• Data Model driven:
  • Follow Custom Resource Definition(CRD) models as much as possible.
• Security:
  • Infra-global and infra-local controller may have privileged access to secrets, keys etc.. Shall ensure to protect them by putting them in HW RoT or at least ensure that they are not visible in clear in HDD/SSDs.
• Redundancy: Infra-global controller shall be redundant, especially, if it used to manage multiple sites.
• Performance: 
  • Shall be able to complete the first time installation or patching across multiple servers in a site shall be in minutes < 10 minutes for 10 server site. (May need to ensure that jobs are done in parallel - Multi-threading of infra-local-controller).
  • Shall be able to complete the patching across sites shall be done in <10 minutes for 100 sites.  

Architecture:

Blocks and Modules

Image RemovedImage Added

All the green items are existing open source projects. If they require any enhancements, it is best done in the upstream community.

All the red items are expected to be part of the Akraino BP.  In some cases, the code in various upstream projects can be leveraged.  But, we made them in red color as we don't know at this time to what extent we can use the upstream ASIS. Some guidance

...

• infra-global-controller-K8S :  This is the K8S cluster where infra-global-controller related containers are run.
• infra-local-controller-K8S:  This is the K8S cluster where the infra-local-controller related containers are run.
• application-K8S :  These are K8S clusters where application workloads are run.

Infra-local-controller: 

"infra-local-controller" is expected to run in bootstrap machine of each location.  Bootstrap is the one which installs the required software in compute nodes used for future workloads.  Just an example, say a location has 10 servers. 1 server can be used as bootstrap machine and all other 9 servers can be used compute nodes for running workloads.  Bootstrap machine is not only installs all required software in the compute nodes, but also is expected to patch and update compute nodes with newer patched versions of the software.

As you see above in the picture, bootstrap machine itself is based on K8S.  Note that this K8S is different from the K8S that gets installed in compute nodes.  That is, these are two are different K8S clusters. In case of bootstrap machine, it itself is complete K8S cluster with one node that has both master and minion software combined.  All the components of infra-local-controller (such as BPA,    Metal3 Metal3 and Ironic) themselves are containers.  

...

Note that infra-local-controller can be run without infra-global-controller. In interim release, we expect that only infra-local-controller is supported.  infra-global-controller is targeted for final Akraino R2 release. it It is the goal that any operations done in interim release on infra-local-controller manually are automated by infra-global-controller. And hence the interface provided by infra-local-controller is flexible to support both manual actions as well as automated actions. 

As indicated above, infra-local-controller is expected to bring K8S cluster on the compute nodes used for workloads.  Bringing up workload K8S cluster normally requires the following steps

1. Bring up Linux operating system.
2. Provision the software with right configuration
3. Bring up basic Kubernetes components (such as Kubelet, Docker, kubectl, kubeadm etc..
4. Bring up components that can be installed using kubectl.

...

1. Select a machine in the location for bootstrapping.
2. Install Linux OS
3. Install Kubernetes on this machine using Kubeadm or any of your favorite tool
4. Upload all binary packages, Linux OSes to be installed in compute nodes using for applications.
5. Upload site specific information - Compute nodes, their roles etc...
6. Once Linux get installed, Via Kuberctl to BPA (via CR), make BPA install the binary packages (such as Kubelet, docker, kubectl, kubenetes API server for application-K8S)
7. Via Kuberctl to BPA, get hold of kubeconfig of application-K8S 
8. Using this kubeconfig, via kubectl to application-K8S, install the packages that can be done via kubectl (such as Multus, OVN Controllers, Virtlet etc...)
9. Make a USB bootable disk for administrators to use in real deployments.
10. Make a VM image for administrators to use in real deployments.

Binary Provisioning Agent (BPA)

BPA job is to install all packages that can't be installed using kubectl to application-K8S.  Hence, BPA is normally used right after compute nodes get installed with Linux operating system, before installing kubernetes based packages.  BPA is also an implementation of CRD controller of infra-local-controller-k8s.  We expect to have following CRs:

• To upload site specific information - compute nodes and their roles
• To instantiate the binary package installation.
• To get hold of application-K8S kubeconfig file.
• Get status of the installation

BPA also provides some RESTful API for doing the following:

• To upload binary images that are used to install the stuff in compute nodes.
• To upload Linux Operating system that are needed in compute nodes.
• Get status of installation of all packages as prescribed before.

...

• When new compute node is added, once the administrator adds new compute node in the site list, it shall take care of installing the packages.
• If a new binary package version is uploaded, it shall take care of figuring out the compute nodes that require this new version and update that compute node with the new version.

...

• KuD that installs basic packages via Kubespray and packages that are not containerized. BPA can inherit this code.
• KuD that acts as private docker hub repository.  BPA can inherit this code.
• KuD that builds the packages from the source code - this needs to be done outside of BPA and binary packages and container packages that result from these are expected to be part of USB bootable disk.
• KuD that brings containerized packages :  This needs to be taken care as a script on top of infra-local-controller.
• CSM  (Certificate and Secret management) can be used ASIS.  Integration with CSM can be for Akraino-R2 and not for interim release

Infra-global-controller: 

There could be multiple edges that need to be brought up.  Administrator going to each location, using infra-local-controller to bring up application-K8S cluster in compute nodes of location is not scalable.  "infra-global-controller" is expected to provide centralized software provisioning and configuration system.  It provides one single-pane-of-glass for administrating the edge locations with respect to infrastructure.   Administration involves

...

Infra-global-controller runs in its own K8S cluster. All the components of infra-global-controllers are containers.  Following containers components are part of the infra-global-controller.

...

• ISTIO and Envoy (for internal communication as well as for external communication) 
• Store Citadel private keys using CSM.
• Store secrets using SMS of CSM.

Admin user experience:

Assuming that infra-global-controller is brought up with all its micro-services, following steps are expected to be taken up to provision sites/edges.

...

Following sections describe the components of infra-global-controller.

Provisioning Controller:

It has following functions

...

• Upon invocation of install, patch or update, it goes through series of steps
  • It uploads the OS images, binary packages  to the site (whatever is needed and if the versions are different from the site)
  • It communicates with  Cluster-API to talk to infra-local-controller Metal3 to prepare it for installation of basic operating system and utilities using Linux image. As part of this, it converts its local inventory information to CRs expected by Cluster-API.
  • Keeps monitoring to ensure the compute nodes are installed with the OS and utilities.
  • Then it communicates with BPM to talk to BPA to install binary packages. 
  • Once BPM provides status that all packages are installed successfully, it gets the kubeconfig of application-K8S.
  • Then it invokes KPM to instantiate containerized packages.
  • Once KPM indicates that all containerized packages are ready/complete, then it provides appropriate status to the user by updating the CR status.

...

• Site registration code can be borrowed from the ONAP K8S plugin service.
• New CRD controller is expected to be created with following CRs
  • Site registration related CRs.
  • Compute inventory related CRs.
  • Site install trigger related CRs.
• Expected to provide APIs
  • For uploading binary packages
  • For uploading containerized packages
  • For uploading OS images
  • Each package, OS image or containerized package is supposed to have right meta data information for identification at later time.

Binary Provisioning Manager (BPM)

K8S Provisioning Manager (KPM)

Design Details

It has following functions

Binary package installation and configuration management :  This functionality via RESTful API is called by PC to trigger the installation.  It then internally calls BPA of infra-local-controller

...

BPA

KuD Changes

Sequence Diagrams

Global ZTP:

Global ZTP system is used for Infrastructure provisioning and configuration in ICN family. It is subdivided into 3 deployments Cluster-API, KuD and ONAP on K8s. 

Cluster-API & Baremetal Operator

One of the major challenges to cloud admin managing multiple clusters in different edge location is coordinate control plane of each cluster configuration remotely, managing patches and updates/upgrades across multiple machines. Cluster-API provides declarative APIs to represent clusters and machines inside a cluster.  Cluster-API provides the abstraction for various common logic that can be seen in various cluster provider such as GKE, AWS, Vsphere. Cluster-API consolidated all those logic provide abstractions for all those logic functions such as grouping machines for the upgrade, autoscaling mechanism.

In ICN family stack, Cluster-API Baremetal provider is metal3 Baremetal Operator, it is used as a machine actuator that uses Ironic to provide k8s API to manage the physical servers that also run Kubernetes clusters on bare metal host. Cluster-API manages the kubernetes control plane through cluster CRD, and Kubernetes node(host machine) through machine CRDs, Machineset CRDs and MachineDeployment CRDS. It also has an autoscaler mechanism that checks the Machineset CRD that is similar to the analogy of K8s replica set and MachineDeployment CRD similar to the analogy of K8s Deployment. MachineDeployment CRDs are used to update/upgrade of software drivers in 

Cluster-API provider with Baremetal operator is used to provision physical server, and initiate the kubernetes cluster with user configuration 

KuD

Kubernetes deployer(KUD) in ONAP can be reused to deploy the K8s App components(as shown in fig. II), NFV Specific components and NFVi SDN controller in the edge cluster. In R2 release KuD will be used to deploy the K8s addon such as Prometheus, Rook, Virlet, OVN, NFD, and Intel device plugins in the edge location(as shown in figure I). In R3 release, KuD will be evolved as "ICN Operator" to install all K8s addons.

ONAP on K8s

One of the Kubernetes clusters with high availability, which is provisioned and configured by Cluster-API will be used to deploy ONAP on K8s. ICN family uses ONAP Operations Manager(OOM) to deploy ONAP installation. OOM provides a set of helm chart to be used to install ONAP on a K8s cluster. ICN family will create OOM installation and automate the ONAP installation once a kubernetes cluster is configured by cluster-API

ONAP Block and Modules:

ONAP will be the Service Orchestration Engine in ICN family and is responsible for the VNF life cycle management, tenant management and Tenant resource quota allocation and managing Resource Orchestration engine(ROE) to schedule VNF workloads with Multi-site scheduler awareness and Hardware Platform abstraction(HPA). Required an Akraino dashboard that sits on the top of ONAP to deploy the VNFs

Kubernetes  Block and Modules:

Kubernetes will be the Resource Orchestration Engine in ICN family to manage Network, Storage and Compute resource for the VNF application. ICN family will be using multiple container runtimes as Virtlet, Kata container, Kubevirt and gVisor. Each release supports different container runtimes that are focused on use cases. 

Kubernetes module is divided into 3 groups - K8s App components, NFV specific components and NFVi SDN controller components, all these components will be installed using KuD addons

K8s App components: This block has k8s storage plugins, container runtime, OVN for networking, Service proxy and Prometheus for monitoring, and responsible application management

NFV Specific components: This block is responsible for k8s compute management to support both software and hardware acceleration(include network acceleration) with CPU pinning and Device plugins such as QAT, FPGA, SRIOV & GPU.

SDN Controller components: This block is responsible for managing SDN controller and to provide additional features such as Service Function chaining(SFC) and Network Route manager.  

Apps/ Use cases:

• SDWAN usecase
• Distributed Analytics as a Service
• EdgeXFoundry use case
• VR 360 streaming

ICN Infrastructure layout

Flows & Sequence Diagrams

Image Removed

1. Use Clusterctl command to create the cluster for the cluster-api-provider-baremetal provider. For this step, we required KuD to provide a cluster and run the machine controller and cluster controller
2. Users Machine CRD and Cluster CRD in configured to instated 4 clusters as #0, #1, #2, #3
3. Automation script for OOM deployment is trigged to deploy ONAP on cluster #0
4. KuD addons script in trigger in all edge location to deploy K8s App components, NFV Specific and NFVi SDN controller
5. Subscriber or Operator requires to deploy the VNF workload such as SDWAN in Service Orchestration
6. ONAP should place the workload in the edge location based on Multi-site scheduling and K8s HPA

Software components

...

Components

...

Link

...

Akraino Release target

...

Cluster-API

...

to initiate the installation and configuration process. It triggers BPA via infra-local-K8S BPA CRs.

Binary package distribution :  This functionality via RESTful API is called by PC.  It figures out the differences between the binary packages & container packages it has locally for this location with the packages that are already in the BPA. Any differences are uploaded to BPA via BPA provided RESTful API.

Collection of KubeConfig of application-K8S : This functionality gets the KubeConfig of application-K8S from BPA. This gets stored in the database table that is specific to site.

K8S Provisioning Manager (KPM)

KPM is used to install containerized packages on application-K8S.  KPM looks at all the relevant helm charts and instantiates them by talking to application-K8S.

Implementation details:

Code can be borrowed from the ONAP Multi-Cloud K8S plugin service which does similar functionality.

Design Details(WIP)

Note : ZTP (Zero Touch Provisioning) term is used in the BP presentation.  This represents both infra-local-controller and infra-global-controller.

Image Added

infra-local-controller

As shown in the above figure, the infra local controller is itself a Bootstrap K8s cluster, that brings up the compute k8s cluster in the edge location.  Infra-local controller has BPA, Metal3, Baremetal operator(Ironic). This section explains the details of it.

Metal3 & Ironic:

This subsection is referred from https://github.com/metal3-io/metal3-docs/blob/master/design/nodes-machines-and-hosts.md

Baremetal operator provides hardware provisioning of compute nodes by using the kubernetes API. The Baremetal operator defines a CRD BaremetalHost Object represents a physical server, it represents several hardware inventories. Ironic is responsible for provisioning the physical servers, and the Baremetal Operator is for responsible for wrapping the Ironic and represents them as CRD object. 

BPA (Define CRD, example CRs, RESTful API)

KuD Changes (Describe how KuD works today and what specific changes would be required)

Metal3 & Ironic

Sequence Diagrams involving all of above + CSM + Logging + Monitoring stuff

Infra-global-controller

PC (Define CRD, Restful API and the example CRs and example API requests)

BPM

KPM

Cluster-API

Global ZTP:

Global ZTP system is used for Infrastructure provisioning and configuration in ICN family. It is subdivided into 3 deployments Cluster-API, KuD and ONAP on K8s. 

Cluster-API & Baremetal Operator

One of the major challenges to cloud admin managing multiple clusters in different edge location is coordinate control plane of each cluster configuration remotely, managing patches and updates/upgrades across multiple machines. Cluster-API provides declarative APIs to represent clusters and machines inside a cluster.  Cluster-API provides the abstraction for various common logic that can be seen in various cluster provider such as GKE, AWS, Vsphere. Cluster-API consolidated all those logic provide abstractions for all those logic functions such as grouping machines for the upgrade, autoscaling mechanism.

In ICN family stack, Cluster-API Baremetal provider is metal3 Baremetal Operator, it is used as a machine actuator that uses Ironic to provide k8s API to manage the physical servers that also run Kubernetes clusters on bare metal host. Cluster-API manages the kubernetes control plane through cluster CRD, and Kubernetes node(host machine) through machine CRDs, Machineset CRDs and MachineDeployment CRDS. It also has an autoscaler mechanism that checks the Machineset CRD that is similar to the analogy of K8s replica set and MachineDeployment CRD similar to the analogy of K8s Deployment. MachineDeployment CRDs are used to update/upgrade of software drivers in 

Cluster-API provider with Baremetal operator is used to provision physical server, and initiate the kubernetes cluster with user configuration 

KuD

Kubernetes deployer(KUD) in ONAP can be reused to deploy the K8s App components(as shown in fig. II), NFV Specific components and NFVi SDN controller in the edge cluster. In R2 release KuD will be used to deploy the K8s addon such as Prometheus, Rook, Virlet, OVN, NFD, and Intel device plugins in the edge location(as shown in figure I). In R3 release, KuD will be evolved as "ICN Operator" to install all K8s addons.

ONAP on K8s

One of the Kubernetes clusters with high availability, which is provisioned and configured by Cluster-API will be used to deploy ONAP on K8s. ICN family uses ONAP Operations Manager(OOM) to deploy ONAP installation. OOM provides a set of helm chart to be used to install ONAP on a K8s cluster. ICN family will create OOM installation and automate the ONAP installation once a kubernetes cluster is configured by cluster-API

ONAP Block and Modules:

ONAP will be the Service Orchestration Engine in ICN family and is responsible for the VNF life cycle management, tenant management and Tenant resource quota allocation and managing Resource Orchestration engine(ROE) to schedule VNF workloads with Multi-site scheduler awareness and Hardware Platform abstraction(HPA). Required an Akraino dashboard that sits on the top of ONAP to deploy the VNFs

Kubernetes  Block and Modules:

Kubernetes will be the Resource Orchestration Engine in ICN family to manage Network, Storage and Compute resource for the VNF application. ICN family will be using multiple container runtimes as Virtlet, Kata container, Kubevirt and gVisor. Each release supports different container runtimes that are focused on use cases. 

Kubernetes module is divided into 3 groups - K8s App components, NFV specific components and NFVi SDN controller components, all these components will be installed using KuD addons

K8s App components: This block has k8s storage plugins, container runtime, OVN for networking, Service proxy and Prometheus for monitoring, and responsible application management

NFV Specific components: This block is responsible for k8s compute management to support both software and hardware acceleration(include network acceleration) with CPU pinning and Device plugins such as QAT, FPGA, SRIOV & GPU.

SDN Controller components: This block is responsible for managing SDN controller and to provide additional features such as Service Function chaining(SFC) and Network Route manager.  

Apps/ Use cases:

• SDWAN usecase
• Distributed Analytics as a Service
• EdgeXFoundry use case
• VR 360 streaming

ICN Infrastructure layout

Flows & Sequence Diagrams

Image Added

1. Use Clusterctl command to create the cluster for the cluster-api-provider-baremetal provider. For this step, we required KuD to provide a cluster and run the machine controller and cluster controller
2. Users Machine CRD and Cluster CRD in configured to instated 4 clusters as #0, #1, #2, #3
3. Automation script for OOM deployment is trigged to deploy ONAP on cluster #0
4. KuD addons script in trigger in all edge location to deploy K8s App components, NFV Specific and NFVi SDN controller
5. Subscriber or Operator requires to deploy the VNF workload such as SDWAN in Service Orchestration
6. ONAP should place the workload in the edge location based on Multi-site scheduling and K8s HPA

Installation demonstration

Software components

Gaps(WIP)

...

Roadmap

Initial implementation (End of July, 2019)

Gaps(WIP)

Roadmap

August Intermediate release

Akraino R2 release

Akraino R3 release

Future releases

Yet to discuss