Is one large sorted set or many small sorted sets more memory performant in Redis

Question

I'm trying to design a data abstraction for Redis using sorted sets. My scenario is that I would either have ~60 million keys in one large sorted set or ~2 million small sorted sets with maybe 10 keys each. In either scenario the functions I would be using are O(log(N)+M), so time complexity isn't a concern. What I am wondering is what are the trade offs in memory impact. Having many sorted sets would allow for more flexibility, but I'm unsure if the cost of memory would become a problem. I know Redis says it now optimizes memory usage for smaller sorted sets, but it's unclear to me by how much and at what size is too big.

Answer 1

如果数据集超过单个主机内存限制，则拥有许多小的排序集将有助于将负载分散到不同的Redis实例上。