Maintaining checksums incurs a small performance overhead. Benchmark testing so far indicates that this overhead is minimal. However, as the impact of the checksum mechanism ultimately depends upon a variety of factors, including disk cache size, cache hits (working set), and the ratio of reads to writes, it varies for each environment.
Sites with tuned and appropriately sized disk caches will experience almost no overhead.
It is important to note that any overhead incurred due to this mechanism will be significantly less than having corrupted blocks going undetected for a long period. The longer the duration a corruption goes undetected, the more time it potentially takes to restore and recover the system and can, in turn, cause more-significant outages.