Multi-server deployment
A multi-server deployment implies that you install the Language Weaver Edge application on one Controller host or machine, while you install the Language Weaver Edge language pairs on several Worker hosts.
Language Weaver Edge multi-server deployment is designed for moderate to high translation volumes. For example, a multi-server deployment can support multiple users translating approximately 100 million words per day, using multiple MT language pairs.
In this multi-server deployment example, Language Weaver Edge is running on three machines: Bergamot, Thymes and Eucalyptus. The Language Weaver Edge application was installed on Bergamot as a Controller host so that it can serve the Web GUI service, which provides a user-interface that is accessible via a web browser, and the REST API service, which provides industry standard HTTP methods for application integrations. Since the usage scenario aims to serve a large pool of simultaneous users, Bergamot is dedicated to hosting the Web GUI and the REST API only, and therefore no Language Weaver Edge language pairs are installed on Bergamot.
Having two Job Engines across two different hosts provide this Language Weaver Edge instance with redundancy and higher capacity for source language detection if the source language is set to auto-detect, for PDF processing, for segmentation of the source input, and for the compilation of the translation into the desired output format. There are more PUs allocated to French to English and Spanish to English translation engines because those languages are expected to have higher usages.
Eucalyptus has a GPU and can host a Training Engine to train Adaptable Language Pairs.
The following are the high level steps taken during the four parts of the deployment process to achieve this Language Weaver Edge cluster setup:
Part 1: Installing the Language Weaver Edge application
The Language Weaver Edge application is installed as a Controller host on Bergamot, while it is installed as a Worker host on Thymes and Eucalyptus.
Part 2: Installing the Language Weaver Edge language pairs
The Language Weaver Edge language pairs are installed on the host machines that will run Translation Engines for those specific language pairs. Specifically, French, Russian, Bengali, Italian, and Japanese to English are installed onThymes. French and Spanish to English are installed on Eucalyptus.
Part 3: Obtaining the Language Weaver Edge license
Only the Language Weaver Edge Controller host requires a license. The Worker hosts get their license from the Controller host.
- The myhosts.json file from Language Weaver Edge instance (retrieved using "Download Entitlement Profile" from in the WebUI).
- The desired Language Weaver Edge language pairs.
- Processing Units count: in this example, 12 PUs are allocated to the different translation engines.
After the license file is provided by the licensing team, it is uploaded through the WebUi.
- Add Thymes and Eucalyptus as hosts.
- Add Job Engines and Translation Engines to each host as the diagram shows.
- If using Adaptable Language Pairs, add a Training engine on a machine with a supported GPU (in this case, Eucalyptus).