Mike said:

>>According to my math your final database size should be on the order
of 100Meg?
>>
>>That means at 200 minutes and 1,000,000 records:
>>83 inserts per second
>>8333 bytes per second

>>Both of these values are terrible.

>>#1 What kind of network connection do you have?  100BaseT?
>>#2 What kind of server are you writing to?
>>#3 How fast does this run if you write to your local machine?

>>Michael D. Black

>It DOES sound terrible since 90%? of the time is spend in retrieving
>data from a remote oracle server over a slow line. The problem is that
>the time spend saving to sqlite is still more than I used to spend on
>saving to my previous file system.


>Can you answer #3 though?

It's nearly 50% faster.

>Does anybody know how to make the journal file go to a different
location than the database?  Apprarently it's not treated as a
"temporary" file.  Perhaps it should be??


Mike said:

Now I'm confused...how can you be 50% faster if 90% of the time is in
retrieving from Oracle?

Werner with a list of processing speeds to clarify matters.

Date       Strt  End   Minutes     recs     Version Filename
2010-05-03 15:42 16:49     66.7 F  421710   4.4.0.6 Due
2010-05-24 01:12 01:30     17.9 F  419604   CLAR    Due
2010-05-27 11:07 13:01    113.5 F  419604   4.4.0.7 Due
2010-06-03 18:54 19:42     47.8 F  419604   4.4.0.9 Due
2010-06-04 17:24 18:09     45.5 B  419604   4.4.0.9 Due
2010-06-07 11:08 13:59    170.5 U  419604   4.4.0.8 Due
2010-06-17 09:13 09:29     16.2 B  419604   CLAR    Due
2010-06-28 13:51 14:07     15.9 F  419604   CLAR    Due
2010-06-28 15:12 15:37     25.4 B  419604   CLAR    Due
2010-06-29 15:57 16:21     24.1 B  419604   CLAR    Due
2010-06-30 11:39 12:17     37.8 F  418620   4.4.1.2 Due
2010-07-08 14:38 15:28     49.4 F  418620   4.4.1.2 Due
2010-07-13 13:09 13:52     42.2 U  418620   4.4.1.2 Due
2010-07-13 18:02 18:31     28.6 F  418620   4.4.1.2 Due (local drive)
2010-07-14 13:27 14:15     47.8 B  418620   4.4.1.2 Due
2010-07-14 15:28 16:03     34.6 F  418620   4.4.1.2 Due

I hope above lines stay in correct columns.
As you can see I started with building these records in 66 minutes.
Where previous version "CLAR" takes between 15 and 24 minutes

After taking out count(*) and adding a few pragma's and saving 6000
records rather than 500 at a time I've got it down to 34 minutes.
If I build in on local drive it takes 28 minutes.(with chunks of 500)

I went with a finecomb thru the code and could not find any other
showstoppers like "count"
Maybe I should dump to a txt file and see what the speed is like?

ps. The last entry (34 minutes) was done with blocks of 6000 rather than
blocks of 500.




DISCLAIMER:
Everything in this email and its attachments relating to the official business 
of Face Technologies is proprietary to Face Technologies. It is confidential, 
legally privileged and protected by law. The person addressed in the email is 
the sole authorised recipient. Any unauthorized dissemination or copying of 
this e-mail (or any attachment to this e-mail) or the wrongful disclosure of 
the information herein contained is prohibited.

_______________________________________________
sqlite-users mailing list
sqlite-users@sqlite.org
http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users

Reply via email to