Re: [GENERAL] Partial indexes Vs standard indexes : Insert performance

2006-08-16 Thread MaXX
Gregory Stark wrote: MaXX <[EMAIL PROTECTED]> writes: In my understanding, a partial index is only touched when a matching row is inserted/updated/deleted (index constraint is true), so if I create a partial index for each protocol, I will slow down my machine as if I had created a single "norma

Re: [GENERAL] Partial indexes Vs standard indexes : Insert performance

2006-08-15 Thread Tom Lane
Gregory Stark <[EMAIL PROTECTED]> writes: > But if you're just looking up a single record I wouldn't expect it to be much > faster to look it up in the smaller partial index than in the larger index. > Indexes find records in log(n) time and log() grows awfully slowly. Yeah. Given the proportions

Re: [GENERAL] Partial indexes Vs standard indexes : Insert performance

2006-08-15 Thread Gregory Stark
MaXX <[EMAIL PROTECTED]> writes: > In my understanding, a partial index is only touched when a matching row is > inserted/updated/deleted (index constraint is true), so if I create a partial > index for each protocol, I will slow down my machine as if I had created a > single "normal" index, but

[GENERAL] Partial indexes Vs standard indexes : Insert performance

2006-08-15 Thread MaXX
Hi, I just want to verify if I'm understanding this correctly: I have a table in which I store log from my firewall. For the protocol column (3 distinct values: TCP ~82%, UDP ~17%, ICMP ~1%, the table contains 1.7M rows), I use a partial index to find ICMP packets faster. In my understanding