From: | "Loic Petit" <tls(dot)wydd(at)free(dot)fr> |
---|---|
To: | scott(at)richrelevance(dot)com |
Cc: | peter(dot)schuller(at)infidyne(dot)com, pgsql-performance(at)postgresql(dot)org, hhall1001(at)reedyriver(dot)com, dforums(at)vieonet(dot)com, max(at)nts(dot)biz(dot)ua |
Subject: | Re: Large number of tables slow insert |
Date: | 2008-08-25 02:02:02 |
Message-ID: | 1775c5ea0808241902u66416c36l935b6df812164678@mail.gmail.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-performance |
That's not a bad idea, at least for historical data.
But actually one of the most common thing in sensor network monitoring is
last readings monitoring.
With indexes what I can do is : SELECT * FROM measures_xx ORDER BY timestamp
DESC LIMIT 1 => And I got the very last reading in a blink (one page reading
only).
It shall be complicated that way... it can only be done by a mass update on
a table each time I receive a packet...
Also it's very important to have last 24h readings aggregate by hour or
minutes (to plot a graph).
So I can't go that way.... I think I must keep the timestamp index where it
is but I should probably get rid of the others.
Thank you again for your help people
Regards
Loic Petit
From | Date | Subject | |
---|---|---|---|
Next Message | Alvaro Herrera | 2008-08-25 02:16:09 | Re: Identifying the nature of blocking I/O |
Previous Message | Tom Lane | 2008-08-25 01:49:51 | Re: Identifying the nature of blocking I/O |