Alexey Pechnikov schreef:
> Hello!
>
> On Sunday 11 October 2009 12:04:37 Ron Arts wrote:
>> CREATE TABLE company(id INTEGER PRIMARY KEY ASC, name)
>>
>> Then I insert 500000 records like this:
>>
>> INSERT INTO company (id, name) VALUES ('1', 'Company name number 1')
>>
>> (with consecutive values for the id value.)
>>
>> do you expect the id column now to be mapped to the Row ID, so
>> this results in the fastest possible way of selecting by id?
>
> Yes, the id is alias for rowid in your table. Check it by
> sqlite3 > explain query plan select name from company where id=1;
>
>> I now get this:
>>
>> sqlite3 prepared insert in trx 500000 records time: 5.08 secs
>> sqlite3 select 500000 records time: 19.28 secs
>> sqlite3 prepared select 500000 records time: 3.47 secs
>> glib2 hash tables insert 500000 records time: 0.37 secs
>> glib2 hash tables lookup 500000 records time: 0.25 secs
>>
>> But I'm still looking to speed up selects.
>
> Hm... I think you may not open more than 8 000 tcp/ip sockets per second
> in common case and so SQLite speed is good enough. Why you write about
> "to handle tens of thousands requests per second"?
>
Well, in my case there can be tens of thousands of connections open
at the same time, where each connection can last days.
Each connection can spit out multiple messages per second, and each
message need around 5 SQL queries.
Ron
> Best regards, Alexey Pechnikov.
> http://pechnikov.tel/
> _______________________________________________
> sqlite-users mailing list
> sqlite-users@sqlite.org
> http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
_______________________________________________
sqlite-users mailing list
sqlite-users@sqlite.org
http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
