Re^6: Threads and DBD::SQLite?

[erix]

Very interesting: I find that, with threads, postgres is not even so much slower. Like you, I am a bit suspicious of the results, but I see no faults in the code.

(All this is with fsync = off (in postgresql.conf), but that (as expected) only seems to influence the writing, not the reading)

So, comparison is again between a SQLite memory table, and a postgresql normal table.

UPDATE I believe there is an oversight in the threading block. ISTM that it doesn't retrieve enough. But I have to admit that I do not quite understand it at the moment. And I'll have to leave it at that for now (bedtime), more later, I hope...

SQLite -  5.96526885 s   Populate db with 100000 records
SQLite -  0.14969087 s   Create primary index
SQLite -  0.54066706 s   Retrieve the whole lot
SQLite -  0.06327486 s   Read all 100000 records using 2 threads
                        
SQLite -  6.06698895 s   Populate db with 100000 records
SQLite -  0.14992595 s   Create primary index
SQLite -  0.53164196 s   Retrieve the whole lot
SQLite -  0.12085700 s   Read all 100000 records using 4 threads
                        
SQLite -  6.14582801 s   Populate db with 100000 records
SQLite -  0.15506721 s   Create primary index
SQLite -  0.55244803 s   Retrieve the whole lot
SQLite -  0.17814016 s   Read all 100000 records using 8 threads
                        
    Pg - 29.35742998 s   Populate db with 100000 records
    Pg -  0.53696704 s   Create primary index
    Pg - 13.75696707 s   Retrieve the whole lot
    Pg -  0.14590406 s   Read all 100000 records using 2 threads
                        
    Pg - 22.90790606 s   Populate db with 100000 records
    Pg -  0.56523991 s   Create primary index
    Pg - 13.84565902 s   Retrieve the whole lot
    Pg -  0.13429594 s   Read all 100000 records using 4 threads
                        
    Pg - 19.43180919 s   Populate db with 100000 records
    Pg -  0.53746796 s   Create primary index
    Pg - 13.00171399 s   Retrieve the whole lot
    Pg -  0.19895601 s   Read all 100000 records using 8 threads

(This is PostgreSQL 9.4devel, btw)