Re: Clusters

From: fumi <fumi__at_tpts5.seed.net.tw>
Date: Fri, 16 Oct 1998 17:03:58 +0800
Message-ID: <707bdq$o0q$1@news.seed.net.tw>

Gocha Mchedlishvili ¼¶¼g©ó¤å³¹ <0lxV1.10$2t4.274_at_news.cwix.com>...
>The first table has about 200,000 rows , the second one about 300,000.
>I built cluster based on PK of the first table and FK of the second one.
>I tried to select 20,000 records.
>Query on clustered tables was 10-15% slower than on ordinary tables

Your SQL statement performs a full table scan of one of the clustered tables,
it will take longer than if the table is unclustered, cause access more blocks required.

Clusters will improve performance if the SQL statement that join clustered tables on the cluster key, by reducing I/O required to retrive data.

Incidentally, INSERT rows into clustered tables or UPDATE cluster key may take longer than a unclustered table, since more overhead is involved in maintaining the cluster key and the index. Received on Fri Oct 16 1998 - 04:03:58 CDT