Why we recommend using DiskChunkCache instead of MemoryChunkCache in production environments？

[mapan1984]

If we have enough memory, for example 64GB, and the original JVM memory allocated to Kafka is only 6G, can I increase the JVM memory to 12G and reserve 6G of memory space for the chunk cache?
In this case, is it possible to achieve higher remote read performance? Are there any disadvantages compared to using 6G of disk space as chunk cache?