each user needs little bit more than 100 000 rows of data. This data is not
filterable and it is atomic.
I have put all the inserts into a transaction it saved 5 mins
with no transaction - 36 mins
with transactin 32 mins
I still need more advices
will it help to use smaller data types?
Aurimas
> One more thing you can try is partitioning your data. Cut the data down to
> what the user will actually use. Does the user need 100,000 rows? Or does
[quoted text clipped - 48 lines]
> | >
> | > Aurimas
Batch the transactions? 5000 at a time?
You could create the DB in the Emulator, which would make use of your
desktop CPU power, and copy to your device.
You could switch to eVC and use OLEDB, which would improve again your
performance about as much as Parameterized queries did the first time.
Use one of SQLCE connectivity scenarios (RDA, Repl) to move the data faster.
--------------------
| From: "Aurimas Norkevicius" <aurimas_norkevicius@studentas.lt>
| References: <uGYF3DerDHA.2772@TK2MSFTNGP12.phx.gbl>
<#e$shXorDHA.4056@TK2MSFTNGP11.phx.gbl>
<t3#izesrDHA.2996@cpmsftngxa07.phx.gbl>
| Subject: Re: very slow inserts
| Date: Thu, 20 Nov 2003 15:06:23 +0200
[quoted text clipped - 7 lines]
| NNTP-Posting-Host: adsl-212-59-23-112.takas.lt 212.59.23.112
| Path:
cpmsftngxa07.phx.gbl!cpmsftngxa10.phx.gbl!TK2MSFTNGXA05.phx.gbl!TK2MSFTNGP08
.phx.gbl!tk2msftngp13.phx.gbl
| Xref: cpmsftngxa07.phx.gbl microsoft.public.sqlserver.ce:10191
| X-Tomcat-NG: microsoft.public.sqlserver.ce
[quoted text clipped - 68 lines]
| > | >
| > | > Aurimas
Christos Chorattides - 22 Nov 2003 07:56 GMT
Do you have indexes on the table?
I haven't tried it on SQLCE but in the old times, when we were doing
mass inserts to a table in a RDBMS, we used to drop all indexes of the
table, do the inserts and re-create the indexes. This was much faster
than inserting thousands of rows into an indexed table.
Aurimas Norkevicius - 25 Nov 2003 11:34 GMT
Yes i have indexes on the table and i use to drop indexes and to recreate
them after insers
> Do you have indexes on the table?
> I haven't tried it on SQLCE but in the old times, when we were doing
> mass inserts to a table in a RDBMS, we used to drop all indexes of the
> table, do the inserts and re-create the indexes. This was much faster
> than inserting thousands of rows into an indexed table.
Christos Chorattides - 22 Nov 2003 08:33 GMT
Do you have indexes on the table?
I haven't tried it on SQLCE but in the old times, when we were doing
mass inserts to a table in a RDBMS, we used to drop all indexes of the
table, do the inserts and re-create the indexes. This was much faster
than inserting thousands of rows into an indexed table.
If you are using supporting development tool you may try filling dataset
with data to insert and
then updating dataset (database_name.UpdateDataSet(..) to perform kind of
bulk insert - this made enormous difference in performance in my case
Rgds
>each user needs little bit more than 100 000 rows of data. This data is not
>filterable and it is atomic.
[quoted text clipped - 15 lines]
>> | >
>> | > Aurimas