Some recommendations
Recommended architecture for development and integration
To develop an application using the FineReader Server COM/Web API, you will need a computer with at least a 4-core CPU, 6 GB RAM and a SDD.
You will be able to install the Server Manager, the Processing Station, the COM API, the Web API, and any other required components all on the same computer if that computer meets the above requirements.
Recommendations for load assessment
- For more accurate load assessment, test-run your workflow as many times as is reasonably possible.
- Test one or two computers with the CPUs, RAM modules, and HDDs that will be used for your workflow. Two computers will be required for testing a distributed system, where documents will be processed on remote Processing Stations.
- If you expect heavy loads, test your workflow on a large queue of several thousand jobs and with all the Processing Stations fully and continuously loaded.
- Check the throughput of your local area network, making sure it can handle the amount of traffic generated by your workflow. The recommended bandwidth is 1 Gbit/s.
- For real-time processing, calculate the number of CPU cores based on the peak load rather than on the average load.
How can I speed up the processing of documents of varying sizes?
If a lot of large multi-page files have to be processed, the processing of one such file may delay the processing of all the other files for a significant period of time. In order to avoid this, the PagesSlice parameter should be set up according to this formula:
PagesSlice = 2 * X * T / N, where
- Х is the maximum allowed number of multi-page files per minute,
- Т is the average recognition time of one such file in a single thread,
- N is the total number of processes.
How can I speed up the processing of office documents (DOCX, XLSX, PPTX, etc.)?
- Use external components (i.e. Microsoft Office or LibreOffice) to take advantage of their multi-threaded code.
- Create a dedicated workflow for office documents and set PagesSlice=0 for that workflow.
How can I improve fault tolerance?
- Consider using Microsoft Cluster Server.
- Alternatively, consider:
- Storing shared data on external RAID disks or in a cloud. In the event of a server hardware failure, you will be able to replace the faulty server with another one (to minimize delay, the back-up server can be prepared in advance)
- Creating regular back-ups of your Processing Stations
Is there a limit to the number of CPU cores supported by FineReader Server?
- FineReader Server has been successfully tested on Processing Stations with 100 CPU cores, without any loss in performance attributable to the Server Manager.
- If required, you can create a multi-server system. You will be able to control several Server Managers from the same Remote Administration Console.
Is there a limit to the number of workflows supported by FineReader Server?
There is no limit to the number of workflows that you can create. However, having more than 100 active workflows in one instance of FineReader Server may slow down the Remote Administration Console. If you require more than 100 active workflows, consider splitting them between two FineReader Server instances.
In the event of a system malfunction, will I need to rescan or reload any unfinished jobs?
- In the event of a Server Manager failure:
- If no operational data has been lost (see "Working with Clusters"), the new Server Manager will pick up where the failed server left off. Only those jobs will need to be rescanned or reloaded that were in progress at the time of the failure.
- In the event of a Processing Station failure:
- The job that was being processed on the failed Processing Station will be automatically redirected to the first available Processing Station (FineReader Server polls the Processing Stations for availability once per second).
Are there any specific requirements for input files (i.e. their maximum size, maximum number of pages in each file, etc.)?
- Maximum image size is 32,000 * 32,000 pixels.
- There is no limit on the number of pages in a file.
- There is no limit on the size of one file.
26.03.2024 13:49:49