Decentralized deduplication operations in a computer system employ a hash index that is a variant of a B+ tree to support both efficient sequential updates as well as efficient random updates. Sequential update is selected when deduplication is infrequently performed, such as on the order of days,