In today's Internet age, stability is a key element for all systems and applications. Whether it is a large data center or a personal computer, stability is the basis for ensuring normal operation and user experience. In this context, let's talk about Milvus, an open source vector database, which introduced the concept of multiple copies of memory in version 2.1.0.
Milvus is an open source vector database mainly used for large-scale similarity search and vector operations. It provides powerful parallel processing capabilities and can support large-scale data queries and operations. In version 2.1.0, Milvus introduced a new feature, memory multiple copies. This feature allows users to increase the QPS (query rate per second) of the system through multiple copies.
Multiple copies of memory are a common method to improve system stability and availability. By creating multiple copies of data on different physical nodes, you can ensure the normal operation of the service even if a node fails or goes offline. At the same time, through the load balancing algorithm, requests can be evenly distributed to each copy to avoid excessive load on a single node, resulting in system performance degradation.
However, this new load balancing strategy also encounters some problems in practical application. The main problem is that QPS does not increase immediately after adding a copy. This is because adding a replica takes some time to synchronize the data, and in some cases, if the amount of data is too large, it may cause the synchronization process to slow down. In addition, when a node goes offline, it takes a long time for the system to return to stability. This is because after a node goes offline, the load needs to be redistributed, a process that may consume a lot of time and resources. There are some other