Why is only the update_item latency relatively high?

During the normal pressure testing process, there is a high latency when only the update_item operation occurs. Why does this phenomenon occur? The CPU load and TPS are not high.

The update operation is an LWT request. And there are some timeout return messages, but the actual request time is within 200ms. (These are known phenomena).

Service unavailable: exceptions::mutation_write_timeout_exception (Operation timed out for alternator_test.test - received only 3 responses from 4 CL=SERIAL. Info=AcquireCurlHandle:2024-06-21 00:26:27.189(0) Transmission:2024-06-21 00:26:27.396(**206**) responseCode:503

Update_items will heavily update the same partitionion. Will this behavior cause an increase in latency?

Yes, concurrent updates to the same partition will increase latency. Make sure to use the latest version of the drivers, which mitigate this somewhat.

I found that when there is put traffic, the update latency will increase, and stopping will decrease. What is the situation? I am using Raid5, is it okay with this?