What This Type of Data Looks Like
Typical subjects include process integration, module, equipment, yield, and manufacturing departments of wafer fabs. Daily documents include process flow cards, batch run records, equipment recipes, SPC charts, exception handling sheets, defect maps, and yield reports. Data sources cover MES systems, equipment FDC, defect detection and metrology equipment, wafer maps, process recipe libraries, and experimental data. Update frequencies include per wafer batch, real-time, and daily. Data contains special fields such as nanoscale process parameters and spatially distributed defect information, with heterogeneous formats and large variations in the size of individual data entries.
What Constraints Do These Characteristics Impose on the Database and Operations Maintenance Link
Multi-source heterogeneous wafer fabrication ledger data requires adaptation to read-write logic for different formats. High-frequency real-time updated equipment data and batch records create continuous write pressure. Ultra-long process chains and nanoscale parameters require databases to have high storage accuracy and low query latency. Data errors can directly lead to wafer batch scrap, so strict transaction guarantees and backup mechanisms are necessary. Equipment data in cleanroom environments cannot be interrupted, requiring database operations maintenance to have rapid recovery capabilities. The limited rework window of the business also requires that the response speed of data query and summarization meets real-time requirements.
Configuration Settings
The following configurations are adapted for FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
| Database connection string | Composite connection configuration adapted for multi-source data | Compatible with ledger formats of different data sources such as MES and equipment FDC |
| Maximum database connections | 60–100 | Handle real-time writes and high-frequency queries per wafer batch, avoid connection exhaustion |
| Redis cache configuration | Calibrated based on actual measurements | Cache frequently accessed wafer maps and yield statistics data to reduce database query pressure |
| Data sharding strategy | Divided by process module | Optimize query latency for ultra-long process chain data, improve cross-module query efficiency |
| Database backup path | /data/fastgpt/wafer_backup | Store daily updated batch run records and exception handling sheets, ensure data recoverability |
| Query timeout setting | 300 seconds | Avoid timeout for cross-module yield attribution queries, align with business processing rhythm |
Three Common Misconfigurations
- Symptom: Knowledge base search response times out, and the number of returned yield attribution results is insufficient. Cause: The maximum database connection count is not adjusted for the high-frequency real-time data of wafer fabrication, leading to database connection exhaustion and failure to respond to query requests in a timely manner.
- Symptom: An insufficient disk space prompt appears when importing wafer batch run records, and a
getaddrinfo EAI_AGAIN mongoerror occurs after restarting the service. Cause: Data storage shards are not divided by process module, single-node disk space is filled with large files such as wafer maps and defect maps, and DNS resolution configuration is not adapted to address changes of multiple database nodes. - Symptom: Parameter format exceptions occur during batch synchronization of equipment FDC data. Cause: A composite connection string adapted for multi-source data is not configured, making it impossible to correctly parse fields and units in different formats from MES and equipment ends.
How to Confirm Proper Configuration
- Execute the database connection test script to verify the connectivity of the database connection string, confirm that the current number of connections does not exceed the configured maximum limit.
- Check the Redis cache hit statistics to confirm that the cache hit rate of frequently queried wafer maps and yield data meets the expected business threshold.
- Import a batch of simulated wafer batch records to check whether the latency of data writing and query meets business requirements.
- Trigger a cross-module yield attribution query to confirm that the returned fields and units are consistent with the standard format of wafer fabrication ledgers.
The values provided are common starting points and should be measured against the reader's own samples.