Sunday, February 19, 2012

Full-text population very slow

Hello, I have a vague question. I have a table with two full-text indexed
columns and 500 000 rows. Previously when only one of the columns was
full-text indexed, the full population took only minutes. Now that there are
two columns, it seems to never finish, or at least is very slow. I have
searched for solution but found nothing, except that perhaps the November CTP
version of SQL 2005 could fix this problem. Does that version fix any known
bugs?
Any help much appreciated...
Thanks, Harri
It is strange. My table is:
CREATE TABLE [dbo].[Search](
[Customer] [varchar](20) NOT NULL,
[Modified] [datetime] NOT NULL DEFAULT ('1900-01-01 00:00:00.000'),
[RowVer] [timestamp] NOT NULL,
[Search] [nvarchar](3999) NOT NULL DEFAULT (''),
[Search2] [ntext] NULL,
CONSTRAINT [PK_Search] PRIMARY KEY CLUSTERED ([Customer] ASC)
)
If I full-text index only Search column, it succeeds in 4 minutes. If I
full-text index only Search2 column, and if all the values in that column are
NULL (but there are over 500 000 rows), then population never finishes.
CREATE FULLTEXT INDEX ON dbo.Search(Search2 LANGUAGE 0) KEY INDEX PK_Search
ON CPMCatalog WITH CHANGE_TRACKING MANUAL
|||Okay, problem solved. It is just the "ntext" data type, it is awfully slow
(same with nvarchar(max)). I guess that I have to create another table, and
remove Search2 column.
|||TEST2
"Harri Pesonen" <HarriPesonen@.discussions.microsoft.com> wrote in message
news:E7A12825-3326-488E-8F47-3F664A6A959E@.microsoft.com...
> Currently I am suspecting that the problem happens because most of the
> rows
> contain same words (this is testing material only). I will next try to
> randomize the words.
|||testyaser
"Harri Pesonen" <HarriPesonen@.discussions.microsoft.com> wrote in message
news:3E84A976-E2DD-4A99-818E-6C923D069E74@.microsoft.com...
> Hello, I have a vague question. I have a table with two full-text indexed
> columns and 500 000 rows. Previously when only one of the columns was
> full-text indexed, the full population took only minutes. Now that there
> are
> two columns, it seems to never finish, or at least is very slow. I have
> searched for solution but found nothing, except that perhaps the November
> CTP
> version of SQL 2005 could fix this problem. Does that version fix any
> known
> bugs?
> Any help much appreciated...
> Thanks, Harri

No comments:

Post a Comment