tag:blogger.com,1999:blog-9149523927864751087.post1483105018057950507..comments2024-03-26T09:43:01.052-07:00Comments on Small Datum: LSM local secondary indexes (LSM LSI)Mark Callaghanhttp://www.blogger.com/profile/09590445221922043181noreply@blogger.comBlogger2125tag:blogger.com,1999:blog-9149523927864751087.post-85586892668307254422020-10-24T10:31:57.012-07:002020-10-24T10:31:57.012-07:00Yes, it is overloaded but distributed databases mi...Yes, it is overloaded but distributed databases might not have been the first use. Used for partitioned tables (local to partition, global across partitions. Then used for distributed databases -- local to shard, global across shards. Now used for LSM -- local to sorted run, global across sorted runs.Mark Callaghanhttps://www.blogger.com/profile/09590445221922043181noreply@blogger.comtag:blogger.com,1999:blog-9149523927864751087.post-78080055770378071872020-10-24T08:51:42.192-07:002020-10-24T08:51:42.192-07:00Good post. It's been a bit of a surprise to me...Good post. It's been a bit of a surprise to me that secondary indexes still seem to be a work in progress for LSM storage engines. So it's a good topic to read and write about.<br />Your definitions for local and global indexes make sense, however I think your names have a risk of being overloaded. In a distributed database I'd expect a global index to be sharded separately from the record it is pointing to. Meaning that a point query only needs to access 1 node in the cluster to read the secondary index.<br /><br />hingohttps://www.blogger.com/profile/09201666166374161923noreply@blogger.com