StarlingX Hardened Managed Kubernetes Platform for the Edge BRENT - - PowerPoint PPT Presentation

StarlingX

BRENT ROWSELL – STARLINGX TSC MEMBER GREG WAINES – STARLINGX CORE BART WENSLEY – STARLINGX CORE STARLINGX.IO

Hardened Managed Kubernetes Platform for the Edge

Agenda

• Architecture Overview
• Deployment Models
• Day 1/Day 2 Operations
• Edge Device Management

StarlingX –Edge Virtualization Platform

StarlingX provides a deployment- ready, scalable, highly reliable Edge infrastructure software platform Services from the StarlingX virtualization platform focus on

• Easy deployment
• Low touch manageability
• Rapid response to events
• Fast recovery

A complete Edge orchestration platform for Bare Metal, VM and Container workloads

* OTHER NAMES AND BRANDS MAY BE CLAIMED AS THE PROPERTY OF OTHERS

MEC == MULTI-ACCESS EDGE COMPUTING

Edge Computing Use Cases

StarlingX –Evolution

• StarlingX Release 1.0 provided

A hardened OpenStack platform on dedicated physical servers

• StarlingX Release 2.0 will provide

A hardened cloud-native kubernetes platform on dedicated physical servers Containerized Openstack Ability to deploy kubernetes only Edge sites for containerized workloads

Physical Servers StarlingX Infrastructure (flock) OpenStack Physical Servers StarlingX Infrastructure (flock) Containerized OpenStack Kubernetes

Container platform

Containerized Application Workloads

Container Platform Details

§ Kubernetes for the orchestration of

container workloads

§ Local replicated docker image registry § Container Runtime: docker § Networking: Calico § Persistent Storage: CEPH RBD § Package Manager

§ HELM § Armada for orchestrating the deployment of

multiple Helm charts

§ Integration with openstack keystone § Local docker image registry authentication § Authentication/authorization of k8s API

Container Platform Details (cont’d)

§ Containerized OpenStack application

§ Leveraging openstack-helm and armada

for deployment

§ Stein based for release 2 § Networking back ends supported: §

Containerized ovs

§

Bare metal ovs-dpdk

§

SR-IOV/PCI-PT

§ Supported services: §

nova, neutron, glance, cinder, glance

§

keystone, barbican, heat, ironic,

§

ceilometer, aodh, panko, gnocchi

StarlingX – Deployment Models

Highly-Available Edge Solution

Two servers

Worker POD Control POD

POD

Storage

Redundant design

Worker

POD

Control

POD POD

Storage Worker

POD

POD POD Worker

POD POD POD

Top of Rack Control Storage Control Storage

Frame-Level Solution

4 –100 servers

Worker POD Control POD POD Storage

Minimum-Footprint Edge Solution

Single server

• Scalable deployment models from 1-100

servers addressing the wide range of edge use cases

• Focus on minimizing the infrastructure footprint

One & two server solution overhead 2 cores/server Frame level solution 2 server master implementation 1 core overhead/worker node

• Storage

Integrated CEPH for one & two node solutions Co-located CEPH on master nodes for small frame level deployments Dedicated CEPH storage nodes for larger configurations

Runs all functions Fully resilient and geographically distributable

Platform Deployment

• StarlingX installs on bare metal servers from an installation

ISO. The ISO can be built using the build tools and instructions found in the StarlingX repos. ISOs are built daily and available from a public repository.

• Docker images for kubernetes components are pulled from

public registries or (optionally) pulled from a private registry.

• Initial system configuration is driven with Ansible:

Modeled specification of target deployment configuration Allows automation of deployments with zero touch provisioning

• Kubernetes components are deployed with kubeadm.

Platform Deployment Procedure

1.

Install initial server from external PXE Boot Server or USB

2.

Run Ansible playbook to configure server as the initial controller / kubernetes master node The playbook can be run locally or from a remote server Deployment specific configuration (e.g. network config) is found in a single bootstrap.yml

3.

Install additional controller, storage and worker nodes from the controller node as follows: Power on server and PXE boot from a controller node -> controller discovers server (receives DHCP request) User selects “personality” (controller, worker, storage) -> controller installs software on server Controller runs inventory discovery on new server to find hardware and resources User configures host details (e.g. interfaces, CPUs, disks, memory assignment) User unlocks server and controller brings it into service

4.

Apply any additional system configuration through configuration APIs

Low touch deployment

OpenStack Deployment

• OpenStack is deployed as a containerized Kubernetes application

OpenStack control plane running in pods OpenStack virtual machines running on host Leverages Kubernetes’ strengths to manage, scale and update the OpenStack services

• Deployed using Helm (using OpenStack-Helm charts) and Armada

(orchestrator for deploying Helm charts from OpenStack Airship).

• StarlingX provides application APIs to install and configure the containerized

OpenStack application Application tarball contains helm charts and armada manifest for StarlingX Automatic generation of helm configuration values based on system configuration User can easily customize helm configuration of OpenStack Services

OpenStack Deployment Procedure

• 1. Apply labels to nodes which will host OpenStack control

plane and compute pods

• 2. Configure local storage for nova on compute nodes
• 3. Import and apply the stx-openstack application
• 4. Configure OpenStack services through the usual

OpenStack APIs

Streamlined OpenStack deployment

Day 2 Operations

• System, Node and Service Configuration Changes,
• System, Node and Service Maintenance:

Administrative commands (e.g. LOCK Node), Fault Monitoring, Alarming and Recovery, Statistics Collection and Analysis, External Centralized Log Management.

• Software Update and Upgrade Management / Orchestration
• Backup & Restore.

Future – Edge Device Management

Support Kubernetes-enabled Devices to JOIN StarlingX Kubernetes Cluster

Devices pre-loaded with Operating System & Kubernetes, StarlingX securely manages ‘kubeadm join ...’

• f Device to StarlingX Kubernetes Cluster.

Value:

• Enables Management of Devices’ Containerized

Workloads from StarlingX Kubernetes Master,

• Enables Devices Apps to access any Services on

StarlingX’s Kubernetes Cluster. Ceph-backed PVCs, Local Docker Registry, etc.

IOT Devices

Future – Device Management (DEMO)

Future – Device Management (DEMO)

What’s next?

• We are actively working on our next release and will be planning the

following release this week. Please feel free to join our PTG session, Thursday noon to Friday EOD !

• Additional StarlingX summit sessions:

Precision Time Protocol (PTP) on StarlingX, Mon 3:50-4:30 Room Level 505/506 Edge Computing Group’s MVP Architecture - StarlingX making it real!, Tue 9:50-10:30 Room Level 505/506 Getting a new Open Source Project off the ground – the StarlingX story, Tue 9:00-9:40 Room Level 505/506 StarlingX for OpenStack Operators, Wed 9:00-9:40 Ballroom Level 4A StarlingX: Hardened Managed Kubernetes Platform for the Edge, Wed 10:50-11:40 Room Level 403/404 StarlingX - Project Onboarding, Wed 11:40-12:00 Room Level 406 StarlingX hands-on workshop, Wed 1:40-5:50 Ballroom Level 4E/F

T

A FULLY FEATURED CLOUD FOR THE DISTRIBUTED EDGE

JOIN THE COMMUNITY

MAILING LISTS: LISTS.STARLINGX.IO FREENODE IRC: #STARLINGX WEBSITE: WWW.STARLINGX.IO JOIN THE FOUNDATION MAILING LIST TO STAY UP TO DATE ON ALL NEW PROJECTS!

Thank You!