1. What is the biggest change in Milvus upgrade from 2.2.x to 2.3.x?
If you can sum it up in one sentence, it means that the usage scenarios are richer. Specifically, it can be reflected from two aspects, namely the deployment environment and the user's sense of use.
For example, from the perspective of the deployment environment, Milvus originally only supported X86 architecture CPUs. After the version upgrade, it can not only support GPUs, but also ARM architecture CPUs. It has been mentioned before, so I won’t go into details here.
As for the user's sense of use, let's take an example. When originally querying, Milvus had certain limitations and could only query more than 10,000 pieces of data. However, the parameters of Milvus 2.3 are configurable. Users who are confident in themselves can adjust the parameters to be larger. In addition, Milvus 2.X version only supports top k. No matter what the distance of the vector is, it will be returned to topk. Now you can filter based on distance and length. In short, there are many similar scenarios, so I won’t list them one by one here.
2. The development cycle of Milvus 2.3.x seems a bit long. Did you encounter any special difficulties during this process?
Indeed, the development cycle of Milvus 2.3.x is relatively long. Initially, the team planned to release it around April, and the functions at that time were almost complete. but,