TechPrufer
Get started
Milvus
Milvus logoMilvus · Feature trail

Consistency

2 posts · Aug 2022 – Sep 2022

What this trail covers

Milvus supports tunable consistency levels (Strong, Bounded, Session, Eventually) by leveraging a time tick mechanism and guarantee timestamps. This allows users to balance data visibility and search latency based on application requirements. The system automatically manages guarantee timestamps based on the chosen consistency level, simplifying user configuration.

All Milvus trails
  1. Understanding Consistency Level in the Milvus Vector Database - Part IIThis post details the underlying time tick mechanism and guarantee timestamp (GuaranteeTs) in Milvus for managing consistency levels. It explains how `ServiceTime` and `GuaranteeTs` interact to determine when search requests can be executed. The post maps the four consistency levels (Strong, Bounded, Session, Eventually) to specific `GuaranteeTs` values and provides examples of how to tune consistency during search and query operations.
  2. Understanding Consistency Level in the Milvus Vector DatabaseThis post explains the concept of data consistency in distributed systems and details the four consistency levels supported by Milvus: Strong, Bounded Staleness, Session, and Eventual. It provides use cases for each level and explains how Milvus 2.0's pub-sub architecture enables these tunable levels, contrasting it with Milvus 1.0's single consistency model. The default consistency level is Bounded Staleness.