IBM Maximo on IBM Cloud
Run a client-managed IBM Maximo Application Suite on Red Hat OpenShift on IBM Cloud (ROKS) in a VPC, or on your own hosts through IBM Cloud Satellite, with you in control of upgrades, databases, and customization.
Overview
Red Hat OpenShift on IBM Cloud, often called ROKS, gives you an OpenShift cluster whose control plane IBM runs, with worker nodes in an IBM Cloud VPC. Maximo Application Suite installs on it the same way it installs on any supported OpenShift cluster, using the MAS CLI or the ibm.mas_devops Ansible collection, and the MAS automation treats ROKS on VPC as the recommended IBM Cloud cluster type. The MAS CLI guide asks for at least 3 worker nodes and recommends no worker smaller than 8 vCPUs and 32 GB of memory. Storage comes from Block Storage for VPC for ReadWriteOnce volumes and File Storage for VPC for the ReadWriteMany volumes MAS needs. IBM Cloud Satellite extends the same managed OpenShift to hosts in your own data center, plant, or another cloud. This is different from IBM's own MAS SaaS and MAS Dedicated offerings, where IBM operates Maximo for you. We size, build, migrate, and operate client-managed MAS on both.
What teams weigh before running Maximo on IBM Cloud
Client-managed on IBM Cloud is not MAS SaaS
Running MAS on your own ROKS or Satellite cluster is a client-managed deployment: you hold cluster-admin, the databases, and the upgrade schedule. IBM's MAS Dedicated is a separate hosting service that IBM documents as offered exclusively on IBM Cloud and run by its SRE team, and IBM's multi-tenant MAS SaaS is described as IBM-managed on AWS. In those services you do not get OS-level access, and customization goes through the UI and APIs. Also note: for IBM Cloud storage, the MAS CLI's automatic detection recognizes the Classic classes, so on a VPC cluster you set the storage classes yourself unless you use OpenShift Data Foundation.
Reference architecture
Red Hat OpenShift on IBM Cloud (ROKS) on VPC infrastructure, with the cluster master managed by IBM. Classic infrastructure is still available, but the MAS Ansible automation labels VPC as the modern, recommended type. ROKS versions use the X.Y_openshift format, for example 4.21_openshift, and must be a version the MAS operator catalog supports.
At least 3 worker nodes for high availability, each at least 8 vCPUs and 32 GB of memory per the MAS CLI guide. VPC flavors use the bx2 family, such as bx2.16x64. Classic flavors such as b3c.16x64.300gb will not work on a VPC cluster. Keep GPU workers in a separate pool so general MAS pods do not land on them.
MAS needs both a ReadWriteOnce and a ReadWriteMany storage class. On VPC, Block Storage for VPC provides RWO volumes and File Storage for VPC provides RWX. Both are zone-specific, so multizone clusters need OpenShift Data Foundation or Portworx if volumes must survive a zone loss. Satellite clusters ship without a storage driver, so you install one.
Maximo Manage needs Db2 or another supported database, MAS core needs MongoDB, and Monitor and IoT need Kafka. Each can run in-cluster through operators or as a separate managed service. The current MAS operator catalog pairs with MongoDB 7.0 to 8.0 and Cloud Pak for Data 5.3.1.
Worker nodes sit on VPC subnets, with a public gateway for outbound pulls from the IBM and Red Hat registries unless you mirror images. On Satellite, no load balancer service is created for the OpenShift Ingress controller, so you plan DNS and load balancing for your hosts yourself.
ROKS on VPC offers GPU worker flavors in the gx2 (V100) and gx3 (L4, L40S) families, availability varying by zone. The MAS CLI can add a GPU worker pool during provisioning for Maximo Visual Inspection and other AI workloads.
Trade-offs
3 minimum worker nodes. MAS CLI guide minimum for a highly available ROKS cluster
8 vCPU / 32 GB smallest recommended worker. The MAS CLI guide does not recommend MAS worker nodes smaller than this
4.18-4.22 openShift versions. Supported by the v9-260924 MAS operator catalog
Teams sometimes assume MAS on IBM Cloud means IBM runs it. On your own ROKS or Satellite cluster, you or your partner own upgrades, patching, backups, and database operations. In MAS SaaS and MAS Dedicated, IBM does, and you give up cluster and OS-level access.
For IBM Cloud storage, the MAS CLI recognizes the Classic classes, not the VPC ones, and it cannot verify that a class supports the access mode you choose. Picking a block class for an RWX volume causes failed or stuck pods.
Block Storage for VPC and File Storage for VPC are tied to one zone. A three-zone cluster gives highly available workers but not highly available volumes, unless you add ODF, Portworx, or database-level replication.
ROKS releases, the OpenShift versions in the MAS operator catalog, Cloud Pak for Data, and MongoDB must line up. ROKS also does not support ImageDigestMirrorSet resources, which affects mirrored or restricted-network installs.
Satellite hosts are yours to supply and keep compliant, and IBM restricts SSH and agent software on them. If the Satellite Link to IBM Cloud drops, running workloads continue, but configuration changes and updates wait until it reconnects.
GPU flavors are not offered in every zone, and the list changes over time. Check ibmcloud ks flavors for your target zone before committing a region for Maximo Visual Inspection.
What we do
We compare client-managed ROKS, Satellite, and IBM-managed MAS SaaS or Dedicated against your customization needs, integrations, data residency, and in-house OpenShift skills, and document what each option means for upgrades and access.
AI on IBM Cloud
We check which AI capabilities you plan to use, such as Visual Inspection, Predict, or the MAS AI Service, and whether they need GPU workers on your platform choice.
We design the VPC, subnets, zones, worker pools, and storage classes, and pick where Db2, MongoDB, and Kafka run. For Satellite, we plan host sizing, the location control plane, the storage driver, and ingress.
AI on IBM Cloud
A dedicated GPU worker pool, sized from IBM's Visual Inspection GPU requirements, keeps training and inference off the general MAS workers.
We provision ROKS and install MAS with the MAS CLI or Ansible collection, with RWO and RWX storage classes set explicitly, then configure Manage, add-ons, SSO, and integrations. Configuration is kept in Git so environments can be rebuilt.
AI on IBM Cloud
We install Maximo Visual Inspection on the GPU pool and confirm the GPU devices are visible to it before handing it to your inspection team.
We move Maximo 7.6.1.x to Manage on IBM Cloud: database upgrade, Java customizations reviewed and repackaged or replaced with automation scripts, integrations retested, and cutover rehearsals against a copy of production.
AI on IBM Cloud
Clean work order and failure history from the migration is what later Health, Predict, and AI Service use cases depend on.
We plan ROKS version updates against MAS catalog support, apply MAS updates and upgrades, and monitor storage, certificates, and backups on a schedule you control.
AI on IBM Cloud
Each upgrade is checked against Visual Inspection model changes, such as the end of SSD model training in MAS 9.1.
Walk through your cluster, databases, integrations, and AI plans with an engineer who installs and runs MAS on OpenShift.
The stack
Requirements
Creating a ROKS cluster for MAS: worker count and size, OpenShift version format, expected storage classes, and the optional GPU worker pool.
Installing MAS on an existing cluster, including the RWO and RWX storage class flags you set on a VPC cluster.
Ansible role for roks and roks-vpc clusters, with VPC flavor, zone, and GPU pool settings.
Which OpenShift, MongoDB, and Cloud Pak for Data versions each MAS catalog release supports.
Block Storage for VPC, File Storage for VPC, Object Storage, ODF, and Portworx, with access modes and zone behavior.
Creating OpenShift clusters in a Satellite location, host requirements, storage, registry, and ingress notes.
IBM's list of GPU devices and requirements for Visual Inspection training and inference.
Yes. ROKS is a managed OpenShift service, and the MAS CLI and MAS DevOps Ansible collection both include provisioning for it. Once the cluster exists, MAS installs the same way it does on other supported OpenShift clusters.
No. A client-managed install on your own ROKS or Satellite cluster gives you cluster-admin, database access, and control of upgrades. MAS Dedicated is an IBM-run hosting service on IBM Cloud, and IBM describes MAS SaaS as IBM-managed on AWS. In those services IBM's SRE team operates the platform and you have no OS-level access.
The MAS Ansible automation calls VPC the modern, recommended ROKS type, and it needs an existing VPC or can create one. The default Classic flavor in that automation fails on VPC, so pick a VPC flavor such as bx2.16x64. Storage classes also differ between the two.
MAS needs a ReadWriteOnce class and a ReadWriteMany class. On VPC, Block Storage for VPC covers RWO and File Storage for VPC covers RWX. For IBM Cloud storage, the MAS CLI auto-detects the Classic ibmc-block-gold and ibmc-file-gold-gid classes, so on VPC you pass the classes yourself unless you run OpenShift Data Foundation, which it also recognizes.
The MAS CLI guide sets a minimum of 3 worker nodes for high availability and does not recommend workers smaller than 8 vCPUs and 32 GB of memory. Actual sizing depends on which applications you install and how many users you have.
Satellite lets you create a location from your own hosts in a data center, at an edge site, or in another cloud, managed by an IBM Cloud region through Satellite Link. You can create Red Hat OpenShift clusters in that location and install MAS on them like any other OpenShift cluster. Check that your Satellite OpenShift version is in the MAS catalog you plan to use.
IBM states that existing workloads keep running. Configuration changes and update rollouts are blocked until connectivity returns, so plan maintenance around the link.
Yes. ROKS on VPC offers GPU flavors such as the gx2 V100 profiles and gx3 profiles with NVIDIA L4 or L40S GPUs, and the MAS CLI can add a GPU worker pool. Confirm the flavor is offered in your zone and is on IBM's supported GPU list for Visual Inspection.
Use a version supported by both ROKS and your MAS operator catalog. The v9-260924 catalog supports OpenShift 4.18 through 4.22. ROKS version names use the X.Y_openshift format.
MAS can use databases running in the cluster through operators or external databases. IBM's upgrade documentation refers to cloud-hosted MongoDB such as on IBM Cloud and points you to that provider's documentation. Check the MAS system requirements for supported database versions first.
Tell us where Maximo runs today and where you want it to run. We will tell you honestly whether IBM Cloud is the right fit.
