"günter" <guenter.kilian@csc.at> wrote in message
news:1074262616.792562@ftp.csc.at...
> hej hej!
>
> We have to build about 10 different FTS on different tabels with up to
> 5.000.000 rows each on SQL 2000 Enterprise Edition - SP3 (Win2000 cluster
> environment).
> Because of an insufficient disc storrage array we can only use 10 hours
> during the night to build up the primary full population (after that we
want
> to change to background tracking).
> so I started yesterday with the first table and about 700.000 rows for the
> first full population (we included a timestamp for further incremental
> populations). It lasted 4 hours to build it. After that I inserted
2.000.000
> rows and started an incremental population for the night. The first thing
I
> got confused was that the Status in the Properties dialog box (Enterprise
> Manager) was changing between "idle" and incremental populating
aproximatly
> every second and in the first 30 minutes no change was reportet for "Item
> Count".
> after 8 hours running the "Item Count" was about 1.000.000 and after 11
> hours at 1.150.000 wat is incredible slower then 4 hours for the first
> 700.000 rows.
> After that I was forced to stop the incremental update.
>
> 1) Is there any suggestion for this behavior
> 2) what happens if I start an incremental population now (I guess that
will
> cause a full rebuild of the fulltext index)
> 3) is there a other possibility to fill up the fulltext index after an
> initial full population (Im thinking about to start with the first
2.000.000
> rows during a night, changing to change_tracking and filling up the table
> every 10 seconds with 100 new rows because if the work load for that isn't
> so high it will work also during the daily office hours and needs a
running
> periode about 83 hours)
> 4) if 3) is a possible scenario, what happens if i want to start another
> fulltext for a second table within the same fulltext catalog?
>
> thanks for all hints and answers
>
> cu
>
> Guenter
>
>

"günter" <guenter.kilian@csc.at> wrote in message
news:1074287326.733800@ftp.csc.at...
> hej John!
>
> Thanks for your informations, so some black clouds are cleared now.
> So I restarted the incremental population.
> But for the time for a Full Population of a table with 2.750.000 rows (one
a
> varchar(254) column for the search values) I needed just 11 hours (with a
> little 4 processor Server with 4 GB Ram- and the indexsize on the disc is
> about 1.6 GB) - but with a short word length and just 2 or 3 words in each
> cell - so the bottleneck was the slow disc array and not the server - even
> that the population was not slow at all ;-)
>
> so, thanks a lot
> Regards
>
> Günter
>
> "John Kane" <jt-kane@comcast.net> schrieb im Newsbeitrag
> news:uFoMnZF3DHA.1184@TK2MSFTNGP10.phx.gbl...
> > Günter,
> > First of all, thank you for providing your environment info!
> > 1. Is there any suggestion for this behavior?
> > A. Yes, the FT Catalog properties are updated asynchronously and various
> > times during the FT Indexing (or crawl) process. Specifically, during a
> > "Shadow Merge" and then after the last "Master Merge" at the end of the
FT
> > Indexing process. The more rows that are indexed the longer you need to
> wait
> > for these FT Catalog counters to be updated.
> >
> > 2. what happens if I start an incremental population now (I guess that
> will
> > cause a full rebuild of the fulltext index)?
> > A. Since you're last step was to " forced to stop the incremental
update",
> > once you re-start the MSSearch service (assuming that this service is
> > stopped), the Incremental Population will automatically pick up where it
> > left off. You might want to switch to "Change Tracking" and "Update
Index
> in
> > Background"
> >
> > 3. is there a other possibility to fill up the fulltext index after an
> > initial full population ?
> > A. Yes, "Change Tracking" and "Update Index in Background". Note, when
you
> > set CT with UIiB, SQL Server will automatically start an Incremental
> > Population on a populated FT Catalog (assuming a timestamp column) or a
> Full
> > Population if the FT Catalog is un-populated.
> >
> > 4. See 3. For a table of 2.7 million rows, you should allow for between
24
> > and 48 hours for a Full Population to complete depending upon your
> hardware
> > and software configurations. See SQL Server 2000 BOL title "Full-Text
> Search
> > Recommendations" as a starting point as well as on your disc storage
> array,
> > you should ensure that at least 15% of this disk array's disk space is
> > always free disk space and available for use by the MSSearch FT Indexing
> > process.
> >
> > Regards,
> > John
> >
> >
> >
> > "günter" <guenter.kilian@csc.at> wrote in message
> > news:1074262616.792562@ftp.csc.at...
> > > hej hej!
> > >
> > > We have to build about 10 different FTS on different tabels with up to
> > > 5.000.000 rows each on SQL 2000 Enterprise Edition - SP3 (Win2000
> cluster
> > > environment).
> > > Because of an insufficient disc storrage array we can only use 10
hours
> > > during the night to build up the primary full population (after that
we
> > want
> > > to change to background tracking).
> > > so I started yesterday with the first table and about 700.000 rows for
> the
> > > first full population (we included a timestamp for further incremental
> > > populations). It lasted 4 hours to build it. After that I inserted
> > 2.000.000
> > > rows and started an incremental population for the night. The first
> thing
> > I
> > > got confused was that the Status in the Properties dialog box
> (Enterprise
> > > Manager) was changing between "idle" and incremental populating
> > aproximatly
> > > every second and in the first 30 minutes no change was reportet for
> "Item
> > > Count".
> > > after 8 hours running the "Item Count" was about 1.000.000 and after
11
> > > hours at 1.150.000 wat is incredible slower then 4 hours for the first
> > > 700.000 rows.
> > > After that I was forced to stop the incremental update.
> > >
> > > 1) Is there any suggestion for this behavior
> > > 2) what happens if I start an incremental population now (I guess that
> > will
> > > cause a full rebuild of the fulltext index)
> > > 3) is there a other possibility to fill up the fulltext index after an
> > > initial full population (Im thinking about to start with the first
> > 2.000.000
> > > rows during a night, changing to change_tracking and filling up the
> table
> > > every 10 seconds with 100 new rows because if the work load for that
> isn't
> > > so high it will work also during the daily office hours and needs a
> > running
> > > periode about 83 hours)
> > > 4) if 3) is a possible scenario, what happens if i want to start
another
> > > fulltext for a second table within the same fulltext catalog?
> > >
> > > thanks for all hints and answers
> > >
> > > cu
> > >
> > > Guenter
> > >
> > >
> >
> >
>
>

> Günter,
> First of all, thank you for providing your environment info!
> 1. Is there any suggestion for this behavior?
> A. Yes, the FT Catalog properties are updated asynchronously and various
> times during the FT Indexing (or crawl) process. Specifically, during a
> "Shadow Merge" and then after the last "Master Merge" at the end of the FT
> Indexing process. The more rows that are indexed the longer you need to
wait
> for these FT Catalog counters to be updated.
>
> 2. what happens if I start an incremental population now (I guess that
will
> cause a full rebuild of the fulltext index)?
> A. Since you're last step was to " forced to stop the incremental update",
> once you re-start the MSSearch service (assuming that this service is
> stopped), the Incremental Population will automatically pick up where it
> left off. You might want to switch to "Change Tracking" and "Update Index
in
> Background"
>
> 3. is there a other possibility to fill up the fulltext index after an
> initial full population ?
> A. Yes, "Change Tracking" and "Update Index in Background". Note, when you
> set CT with UIiB, SQL Server will automatically start an Incremental
> Population on a populated FT Catalog (assuming a timestamp column) or a
Full
> Population if the FT Catalog is un-populated.
>
> 4. See 3. For a table of 2.7 million rows, you should allow for between 24
> and 48 hours for a Full Population to complete depending upon your
hardware
> and software configurations. See SQL Server 2000 BOL title "Full-Text
Search
> Recommendations" as a starting point as well as on your disc storage
array,
> you should ensure that at least 15% of this disk array's disk space is
> always free disk space and available for use by the MSSearch FT Indexing
> process.
>
> Regards,
> John
>
>
>
> "günter" <guenter.kilian@csc.at> wrote in message
> news:1074262616.792562@ftp.csc.at...
> > hej hej!
> >
> > We have to build about 10 different FTS on different tabels with up to
> > 5.000.000 rows each on SQL 2000 Enterprise Edition - SP3 (Win2000
cluster
> > environment).
> > Because of an insufficient disc storrage array we can only use 10 hours
> > during the night to build up the primary full population (after that we
> want
> > to change to background tracking).
> > so I started yesterday with the first table and about 700.000 rows for
the
> > first full population (we included a timestamp for further incremental
> > populations). It lasted 4 hours to build it. After that I inserted
> 2.000.000
> > rows and started an incremental population for the night. The first
thing
> I
> > got confused was that the Status in the Properties dialog box
(Enterprise
> > Manager) was changing between "idle" and incremental populating
> aproximatly
> > every second and in the first 30 minutes no change was reportet for
"Item
> > Count".
> > after 8 hours running the "Item Count" was about 1.000.000 and after 11
> > hours at 1.150.000 wat is incredible slower then 4 hours for the first
> > 700.000 rows.
> > After that I was forced to stop the incremental update.
> >
> > 1) Is there any suggestion for this behavior
> > 2) what happens if I start an incremental population now (I guess that
> will
> > cause a full rebuild of the fulltext index)
> > 3) is there a other possibility to fill up the fulltext index after an
> > initial full population (Im thinking about to start with the first
> 2.000.000
> > rows during a night, changing to change_tracking and filling up the
table
> > every 10 seconds with 100 new rows because if the work load for that
isn't
> > so high it will work also during the daily office hours and needs a
> running
> > periode about 83 hours)
> > 4) if 3) is a possible scenario, what happens if i want to start another
> > fulltext for a second table within the same fulltext catalog?
> >
> > thanks for all hints and answers
> >
> > cu
> >
> > Guenter
> >
> >
>
>