Light weight solution for distrubted memory map
-
Hi Our project uses a stl map to save meta info. The map is the entrance of all operations, include read / write / update, etc. It must be very efficient. Every item of the map is about 0.1kB. Now the length is 1M at most. So, the memory used is not over 100M. It is OK by now. However, in the further, the length might be 1G or more. We have to consider a distrusted solution with high performance… We tested memcached / Redis. The performance is not so good (no more than 5k query per seocns, need time to optimise.) What’s more, both the solution (memcached / redis) bring many issues for product deployment. Is there a light weight solution for distrusted map in memory? For example, hold 3 maps in 3 servers and a meta in primary. In detail, 1. All query / update / insert / delete operation is finished in primary 2. If the info is in primary, do it locally, otherwise, send request (query / insert / update / delete) to peer. 3. Three parts of information can be balanced dynamically. Thanks Jack