Data in caching systems tends to follow a write-once, read-many access pattern, so a higher compression level can be applied at write time while relying on high-speed decompression at read time to guarantee response latency.