From: | Leonardo Francalanci <m_lists(at)yahoo(dot)it> |
---|---|
To: | pgsql-hackers(at)postgresql(dot)org |
Subject: | Re: Fast insertion indexes: why no developments |
Date: | 2013-11-05 15:22:15 |
Message-ID: | 1383664935793-5777020.post@n5.nabble.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
Claudio Freire wrote
> real data isn't truly random
Well, let's try normal_rand???
create table t1 as select trunc(normal_rand(1000000, 500000, 30000)) as n,
generate_series(1, 1000000) as i;
with cte as
(select min(n) as minn, max(n) as maxn, i/100 from t1 group by i/100),
inp as (select generate_series(1, 100) iinp, trunc(normal_rand(100,
500000, 30000)) as s)
select count(*),iinp from inp
left outer join cte on inp.s between minn and maxn group by iinp;
Not that much different in my run...
Claudio Freire wrote
> you haven't really
> analyzed update cost, which is what we were talking about in that last
> post.
I don't care for a better update cost if the cost to query is a table scan.
Otherwise, I'll just claim that no index at all is even better than minmax:
0 update cost, pretty much same query time.
Maybe there's value in minmax indexes for sequential data, but not for
random data, which is the topic of this thread.
BTW I would like to see some performance tests on the minmax indexes
vs btree for the sequential inputs... is the gain worth it? I couldn't find
any mention of performance tests in the minmax threads.
--
View this message in context: http://postgresql.1045698.n5.nabble.com/Fast-insertion-indexes-why-no-developments-tp5776227p5777020.html
Sent from the PostgreSQL - hackers mailing list archive at Nabble.com.
From | Date | Subject | |
---|---|---|---|
Next Message | Stephen Frost | 2013-11-05 15:32:25 | Re: [PATCH] configure: add git describe output to PG_VERSION when building a git tree |
Previous Message | Andres Freund | 2013-11-05 15:21:41 | Re: logical changeset generation v6.5 |