Is there any benefits of making multiple ConnectionMultiplexer instances when using StackExchange.Redis? We are making heavy read/write call to Azure Redis Cache and wondering how much load can single ConnetionMultiplexer can handle.
Currently we have a pool of ConnectionMultiplers in array format and pick one randomly to handle concurrent calls. If single ConnectionMultiplexer can do the job, then tis is unnecessary implement.
Only sometimes. My experience is that apps which run at the right scale, on the right kind of hardware, with the right kind of load to need multiple connection multiplexers (for a single redis cache) are few and far between.
More specifically, most apps get by just fine with just one ConnectionMultiplexer. But I have seen a couple cases where you might be better off using num_cpus / 4. These are generally apps with not too many client machines (e.g. < 1000), and each client machine is fairly powerful (e.g 8+ cores).
One other scenario where you might possibly see some benefit is with transient connection breaks due to packet loss, but you might want to fix your n etwork in that case
