-Milvus is a system that separates storage and computation. In this system, DataNodes are responsible for the persistence of data and ultimately store it in distributed object storage such as MinIO/S3. QueryNodes handle computational tasks like Search. These tasks involve processing both batch data and streaming data. Simply put, batch data can be understood as data that has already been stored in object storage, while streaming data refers to data that has not yet been stored in object storage. Due to network latency, QueryNodes often do not hold the most recent streaming data. Without additional safeguards, performing Search directly on streaming data may result in the loss of many uncommitted data points, affecting the accuracy of search results.
0 commit comments