Playbooks
Deployment Sequence
Recommended rollout sequence for a new BullSequana AI deployment
This playbook describes the recommended sequence for a new BullSequana AI deployment.
It applies across cloud, hybrid, and on-premise environments, with extra hardware-specific steps when BullSequana systems are delivered as part of the project.
Note This page is an enterprise rollout playbook. New technical readers should start with Choose a Deployment Model, Before You Start, and Prerequisites before using this sequence.
1. Assess Requirements And Sizing
| Num. | Activity title | Task description | Input | Output | Teams | Deployment type | |
|---|---|---|---|---|---|---|---|
| 1 | Requirement provisioning | Client specifies the functional requirements for the platform. This task should result in a functional specification document enabling the definition of software and hardware requirements. | Functional requirements | Functional spec. | CoreAI | Cloud, Hybrid, On-prem | |
| 2 | Basic sizing provisioning | Assessment of the requirements like : softwares, firmware, client constraints, ... A technical specification document is then made for the HPC and CoreAI teams. | Technical requirements | Technical spec. | HPC and CoreAI | Cloud, Hybrid, On-prem | |
| Specification validation | Client | Cloud, Hybrid, On-prem |
2. Prepare And Deliver Hardware
This phase mainly applies to BullSequana hardware and other on-premise delivery models.
Note The following sequence is only relevant in the case of an on-premise deployment.
Steps up to shipment typically happen in the factory, while the remaining work happens at the client site.
| Num. | Activity title | Task description | Input | Output | Teams | Deployment type |
|---|---|---|---|---|---|---|
| 3 | Hardware assembly | Assembly of the hardware at the factory | Technical spec. | Assembled hardware | HPC | On-prem |
| 4 | Network requirements | Provision of network addresses and VLAN requirements for the entire solution. This is needed for network QA and AT. | Network requirements | Network configuration on the hardware | HPC | On-prem |
| 5 | Operating system install | Installation of: OS, backup agent, Kubernetes, system container registry and miscellaneous softwares. And technical validation of all components, QA. Note: the Coreai Platform can be installed at this step if requested, refer to the table below for the additional steps | Hardware & Software spec. | Bootable hardware | HPC | On-prem |
| Acceptance testing | This final stage determines if all the hardware and software is compliant with the acceptance criteria determined in the specifications documents. At this stage the hardware should be ready to ship to the client site | HPC | On-prem | |||
| 6 | On-site delivery | The hardware components will be turned off and disconnected to be delivered to the client site. | AT greenlight | Hardware shipment | HPC | On-prem |
| 7 | On-site installation | Hardware installation and tuning of the software configuration | Hardware delivery | Hardware installation | HPC | On-prem |
| 8 | Admin workstation deployment | Deployment of a central admin workstation. It enables the hosting of the system container registry used by Kubernetes. Additionally it will host the deployment scripts for the chosen Kubernetes distribution. | Technical spec. | Admin workstation and system container registry | HPC | On-prem |
| Delivery validation | Client signature of the receipts for both the delivery and installation of the hardware and software. | Installed hardware and software | Signed receipts | HPC | On-prem |
3. Finalize Security And Exposure Inputs
| Num. | Activity title | Task description | Input | Output | Teams | Deployment type |
|---|---|---|---|---|---|---|
| 9 | DNS provisioning | Definition and provisioning of DNS clients for all domain names required for the project (services and infrastructure). | List of services and endpoints for the platform | DNS zone and records | Client or CoreAI | Cloud, Hybrid, On-prem |
| 10 | Certificate provisioning | Inventory of needed certificate and provisioning by the client or by the CoreAI Certificate Authority (CA). These certificates will be self-signed by a single CA (PKI). | DNS records | Certificates and optionally a CA | Client or CoreAI | Cloud, Hybrid, On-prem |
4. Bootstrap The Platform
The following list assumes the functional and technical specification is already available.
| Num. | Activity title | Task description | Input | Output | Teams | Deployment type |
|---|---|---|---|---|---|---|
| 1 | Kubernetes deployment | Kubernetes cluster deployment. The deployment take into account the network infrastructure settings. | Network configuration, Kubernetes distribution, system container registry available | Kubernetes cluster | HPC or CoreAI | Cloud, Hybrid, On-prem |
| 2 | Deployment of additional linux packages | Retrieve linux packages to be independent from internet mirrors. | Technical spec. | Package repository | HPC | On-prem |
| 3 | [Optional] CoreAI container registry installation | A platform registry is prepared and all necessary images are loaded. Note: if the client already operates a compatible registry this step can be skipped, but factory preloading for on-premise delivery depends on having a delivery path agreed in advance. | Technical spec. | Registry and platform images | HPC | Cloud, Hybrid, On-prem |
| 3 - ALT | [Optional] Client container registry init | If the client uses an existing registry, BullSequana AI images are loaded there instead. | Technical spec. | Platform images in target registry | HPC or CoreAI | Cloud, Hybrid, On-prem |
| 4 | [Optional] GitLab deployment | A Git service is prepared to host platform configuration and IaC. Note: if the client already has GitHub or GitLab in place, this step can be replaced by repository initialization only. | Functional spec. | Git service and repositories | CoreAI | Cloud, Hybrid, On-prem |
| 4 - ALT | [Optional] Git repository initialisation | Push of the source code for the CoreAI platform and IaC. | Functional spec. | Git repositories | CoreAI | Cloud, Hybrid, On-prem |
| 5 | Platform deployment | Deployment of the platform scope that was selected for the project, including Runtime, CoreAI, ProAI, and any approved use-case extensions. | Functional spec. | Platform services and IaC | CoreAI | Cloud, Hybrid, On-prem |
| Technical Acceptance testing | AT for the Kubernetes cluster deployment and CoreAI deployment. | Kubernetes access | AT greenlight | Client | Cloud, Hybrid, On-prem |
5. Tailor And Validate
| Num. | Activity title | Task description | Input | Output | Teams | Deployment type |
|---|---|---|---|---|---|---|
| 6 | IDP federation and IAM configuration | The CoreAI IAM component (Keycloak) can be federated with any pre-existing Identity provider. And specific configuration for IAM can be applied if requested. | Technical spec. | IAM configuration | CoreAI | Cloud, Hybrid, On-prem |
| 7 | Generic component configuration | By request of the client the CoreAI team will tune the configuration of the components to satisfy either use cases constraint or technical and security requirements. | Additional technical spec. | Component configuration | CoreAI | Cloud, Hybrid, On-prem |
| Final round of Acceptance testing | The client perform the last round of AT and ensure that the platform is fully tailored to their needs. | Client | Cloud, Hybrid, On-prem |