Skip to main content

Dedicated Computing Power

Starting from version v5.1.0, the creation of dedicated computing power services within an organization is supported. By adding important workflows to these dedicated computing services, you can prevent them from being affected by congestion from other platform processes.

Dedicated computing power services are supported only when deployed in cluster mode. By using the tolerations feature of Kubernetes, the dedicated computing services (workflowcompute) are run on servers isolated from microservices, thus avoiding resource occupation by other services.

Once the dedicated computing power service is deployed, the server creates the workflowcompute service based on the dedicated computing resource specifications selected on the page. This service is named in the format of workflowcompute-ResourceID on Kubernetes and is displayed as a deployment resource. Within each pod, the following three containers will be initiated:

  • istio-proxy: This is the proxy container of Istio.
  • worksheetservice: This is the worksheet service container responsible for handling tasks related to worksheets.
  • workflow-consumer: This is the workflow consumer service container responsible for processing tasks related to workflow consumption.

Moreover, Kubernetes limit rules will restrict CPU and memory resources for each container, with the limits typically matching the dedicated computing resource specifications shown on the page.

Additionally, the system will create a new topic in Kafka based on the selected concurrency level. This topic is named in the format of md-workflow-consumer-ResourceID, with the number of partitions defaulting to the selected dedicated computing concurrency level.

Once specific workflows are added to the dedicated computing services, the corresponding processes will proceed through an independent topic, utilizing dedicated workflow consumer and worksheet services for processing, thereby avoiding congestion impact from other platform processes.