The MIT-licensed CSM codebase mixes deployments with software through a unified installer that only supports a reliatively small set of hardware configurations. In order to use ochami on smaller systems and with other hardware targets, we need a clean separation between the microservices themselves and the infrastructure/deployment technologies that HPE has chosen.
Deployment Targets
Fully Integrated
HPE will maintain the CSM installer and full NCN deployment tooling as part of their contracts to support CSM with existing customers. This unified installers targets a minimum of 9 nodes that are divided into masters, workers, and storage nodes and serve as a common hardware platform for all other Shasta products. This fully integrated system supports full air-gap installation of Kubernetes and Ceph with integrated services that also provide S3-like storage, databases, Kafka-as-a-Service, Software Repositories, and more. All services are deployed in a full HA configuration with microservices scaled to fit the needs of the customer.
Minimal Installation
The ochami project needs a minimal reference implementation of the microservices that can be independently tested from the supporting infrastructure. Partners can start with the minimal installation and build their own custom installations for use with small clusters or for cloud testing.
BYOI Installation
Many of the fully integrated services in the CSM package can be obtained in other ways. HPC sites generally have their own package repositories, logging systems, and services or teams that can provide as-a-service access to Kubernetes, databases, and other infrastructure. The microservices themselves are abstracted from the infrastructure. Ochami is targeting a deployment models that allow for sites to swap in their own infrastructure and other functionality with clean boundaries and conformance tests.
The MIT-licensed CSM codebase mixes deployments with software through a unified installer that only supports a reliatively small set of hardware configurations. In order to use ochami on smaller systems and with other hardware targets, we need a clean separation between the microservices themselves and the infrastructure/deployment technologies that HPE has chosen.
Deployment Targets
Fully Integrated
HPE will maintain the CSM installer and full NCN deployment tooling as part of their contracts to support CSM with existing customers. This unified installers targets a minimum of 9 nodes that are divided into masters, workers, and storage nodes and serve as a common hardware platform for all other Shasta products. This fully integrated system supports full air-gap installation of Kubernetes and Ceph with integrated services that also provide S3-like storage, databases, Kafka-as-a-Service, Software Repositories, and more. All services are deployed in a full HA configuration with microservices scaled to fit the needs of the customer.
Minimal Installation
The ochami project needs a minimal reference implementation of the microservices that can be independently tested from the supporting infrastructure. Partners can start with the minimal installation and build their own custom installations for use with small clusters or for cloud testing.
BYOI Installation
Many of the fully integrated services in the CSM package can be obtained in other ways. HPC sites generally have their own package repositories, logging systems, and services or teams that can provide as-a-service access to Kubernetes, databases, and other infrastructure. The microservices themselves are abstracted from the infrastructure. Ochami is targeting a deployment models that allow for sites to swap in their own infrastructure and other functionality with clean boundaries and conformance tests.