Cluster installation
Determing if a cluster installation of Spotfire Statistics Services is the right installation for your needs.
In your plans for your deployment, if your requirements include the scalability, high-availability, and reliability that go beyond the benefits of a standalone installation, consider deploying Spotfire Statistics Services in a clustered environment.
You can install Spotfire Statistics Services Manager and Worker nodes on multiple computers so that they will operate together in a clustered environment. You can also configure a Manager node to fill the role of a Worker node so that it responds to administration requests and processes jobs. In this way a node can function as an entire system on a single computer or as many computers participating in a cluster.
Cluster Manager node
You must have at least one Manager node in each cluster, and it must be installed on a computer separate from the Worker nodes. When you install additional Manager or Worker nodes the configuration settings from this first Manager node is read in during the installation process.
The role of a Manager node is to expose a set of RESTful resources on the network that provide the ability to submit an expression or function for execution to an engine, enable remote job management, and configure and manage a cluster.
The Manager node also exposes a WebDAV service for remote file management. This service allows clients to upload a data set to a server prior to job execution and then, once the job is complete, they can obtain the files generated as an output of their job execution.
The Manager node has the following components.
| Manager node component | Description |
|---|---|
| HTTP Request Handler | Translates HTTP requests into processing runtime or cluster management component calls. |
| Processing Runtime | Queues jobs and executes job requests. |
| WebDAV Service | Provides server directory and file access. |
| Cluster Management | Manages all of the nodes in a cluster and distributes load between nodes. |
Cluster Worker node
The Worker node is a command-line application wrapped in a service that listens for commands from Manager nodes and processes queued jobs. A Worker node does not expose any externally accessible services other than those available through a Manager node.
The Worker node has the following components.
| Worker node component | Description |
|---|---|
| Processing runtime | Queues jobs and executes job requests. |
| Cluster Agent | Listens for commands from the Manager node; Starts, stops, or restarts the processing runtime as needed; and returns job status change notifications. |
For more information, see Cluster deployment.