Re: Large range query stops using indexes | Thread profile

How are index scans a "random read"? Traditionally, indexes are normally B-Trees, unless specified otherwise. (I'm pretty sure that my index in question is a B-Tree). ...of which, I envision the query should an index for range scans as follows: Locate low value, locate high value, parse all in-between branches and return all leaves. It seems not to be doing that. Sigh.

Correct, it gets the left most node and the right most node and scan everything in between. However, unlees it is the clustered index, it will have to use the pointer or whatever to read the row having that value in the node in the b-tree. So, there will be a random scan to read the rows that satisfy the range. If the left-most and right-most are near to the left and right end of the tree, meaning it needs to scan huge percentage of the tree,...