2 ● Why have support for different architectures? ○ OpenShift adoption in IBM Power and IBM Z environments ○ The rise of Arm: cloud adoption and servers ○ The customer need for flexibility ● What about the drive for mixed architecture clusters? ○ What are the main use cases? ○ What are the customer benefits? ○ Are there different ways to solve the same problem? ○ Is this the way the industry is going? OPENSHIFT EVERYWHERE Let’s talk about multiple architectures
3 OpenShift on Arm ● Run OpenShift on highly efficient, high performance per watt architectures o-----------------------------o ● Support for Arm on Azure (UPI) ● Single Node OpenShift on Arm (bare metal) Multi-architecture Cluster ● Allow more flexibility in a cluster, use different cloud platforms o------------------------------o ● Multi-architecture compute platforms: ○ AWS Arm support (GA) ○ Azure Arm support (GA) ○ Bare Metal Arm (TP) ● Multi-architecture compute migration and upgrade support ● Hosted Control Plane: ○ AWS Arm control plane IBM Power and zSystems ● Run OpenShift on highly available, highly secure, scalable hardware o-----------------------------o ● FIPS Mode supported ● Cluster Resource Override Operator ● Network Bound Disk Encryption ● Metal LB support ● Egress IP support PM: Duncan Hardie Systems Enablement - latest highlights
V0000000 Arm, Driving the Ecosystem 4 Reasons for Considering Arm ● Performance per Watt ● Cloud adoption with competitive price/performance ● Edge adoption ● Server support ● Rapidly growing ecosystem ● And finally … the same OpenShift experience you are used to today OPENSHIFT EVERYWHERE
5 OPENSHIFT EVERYWHERE OpenShift on Arm Full Stack Automation (IPI) Bare Metal Pre-existing Infrastructure (UPI) Bare Metal Layered Product Support OpenShift GitOps Red Hat Advanced Cluster Manager Logging OpenShift Pipelines
Heterogeneous Cluster Use Cases 7 OPENSHIFT EVERYWHERE Control plane Compute nodes Control plane Compute nodes Legacy Applications Cost Optimization Architecture Migration ● Deal with that pesky app that you don’t have the source code for and is too expensive to re-write ● Handle gaps in the new architecture ecosystem ● Use a different architecture control plane that is made up of “cheaper” systems ● Interesting to cloud providers and also IBM zSystem and IBM Power ● Easy way to try out specific apps/services on a different architecture while maintaining your environment ● Gradual roll out of change to another architecture not requiring “big bang” change everything at once approach ● Interesting benchmark opportunity Control plane Compute nodes
Multi-architecture Compute - Adding new nodes 8 Control plane Compute nodes x86 Arm Control plane Compute nodes Control plane Compute nodes 1 Create/Install your OpenShift Cluster as normal 2 Add in different architecture nodes as a day 2 operation 3 Schedule workloads on your heterogeneous cluster OPENSHIFT EVERYWHERE
V0000000 Demo time: Multi-Architecture Compute Machines 9 EVOLUTION OF ARCHITECTURE Control plane Compute nodes x86 Arm 1 Create/Install your OpenShift Cluster as normal 3 Configure your MachineSet (AWS) 4 Apply MachineSet and watch Arm node(s) get added x86 Arm 2 Point to multi-payload and update x86 payload Control plane Compute nodes x86 payload multi payload x86 Arm Control plane Compute nodes
V0000000 Multi-Architecture Compute Functionality Supported at GA Platform Support at GA ● Installation of a OpenShift cluster using a manifest-listed release payload ● Adding compute nodes of a different architecture as a day 2 operation ● Automated upgrades between multi-arch releases ● Migration support through oc commands to migrate from a single arch to a multi-arch release payload ● Imagestreams support importing manifestlisted images and the internal registry also supports local pull through. ● oc commands like `oc tag`, etc…also have support to tag/import manifestlists OPENSHIFT EVERYWHERE Bare Metal (Tech Preview)
V0000000 Multi-Architecture Compute Current Limitations ● Out of the box Multi-Arch aware scheduling ● Day 0 deployments with multiple architecture compute ○ For installation all nodes need to be of a single architecture. ● Disconnected installations ○ Not tested yet but can be done with some manual steps ● Autoscaling ○ support for arch nodes based on weights (e.g “prefer” to use Arm instances) ○ Autoscaling from zero ● OpenShift builds classic (build v1) don’t support creating manifest listed images ○ Use OpenShift Pipelines as a workaround OPENSHIFT EVERYWHERE
12 Low CAPEX and OPEX costs (bundling of CPs + CP as pods) Central Management of CPs (easy operation & maintenance) Multi-arch support (e.g. x86 and Arm) Network & Trust segmentation Mixed Iaas For CP and Workers Fast cluster bootstrapping (CP as Pods) OPENSHIFT EVERYWHERE Hosted Control Planes (Control planes as pods) Hosted Control Planes and Multi-architecture compute
linkedin.com/company/red-hat youtube.com/user/RedHatVideos facebook.com/redhatinc twitter.com/RedHat Thank you Red Hat is the world’s leading provider of enterprise open source software solutions. Award-winning support, training, and consulting services make Red Hat a trusted adviser to the Fortune 500.
redhatlivestreaming
s2terminal
onecareer_tech
masasuzu
muziyoshiz
cimadai
ufried
niftycorp
dahatake
is_ryo
h1kita
murachiakira
nwiizo
bkeepers
tenderlove
addyosmani
revolveconf
colly
michaelherold
dougneiner
davidbonilla
jennybc
mongodb
hatefulcrawdad