News for hackers. Stuff that matters.
Re:Understanding Postgres Performance by lmm on Tuesday October 02, 2012 @02:34PM (#4603486)
If n is your number of rows, adding a constant cost to each write is an O(n) total penalty. But the asymptotic performance isn't really the point for many real-world workloads (e.g. an O(1) hashtable lookup is generally much slower than an O(n) linked list traversal). As the grandparent says, read performance problems are generally easier to work around than write performance ones, and reporting aggregation reads (which are the ones that tend to require table scans) are often far less time-sensitive than writes.

There are no comments on this comment yet.