You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository has been archived by the owner on Jan 13, 2022. It is now read-only.
Hi, we have been testing MonetDBJavaLite as an alternative to "normal" monetDB and as expected we see a significant improvement when retrieving batch data (i.e. an entire column).
However we found some rather big perfomance drops when using queries like ("Select * from table where value between 100 and 300") in larger tables.
These are the result for 10 million rows and as we can see the normal monetdb becomes significantly faster following the first query while the Monetdbjavalite only have minor difference even when manually adding indexing to the table.
It seems like Monetdbjavalite are either missing some initial query optimizations that normal monetdb are using or are repeating them for each new query.
Test for Monet non embedded with 10'000'000 rows
Data generation took 3932ms
Copy binary took 1683ms
First query took 999ms
Second query took 78ms
Third query took 79ms
Test for Monet Embedded with 10'000'000 rows
Data generation took 4199ms
Copy binary took 2388ms
First query took 1031ms
Second query took 906ms
Third query took 967ms
Test for Monet Embedded with 10'000'000 rows (with explicitly added index)
Data generation took 3853ms
Copy binary took 2512ms
First query took 1092ms
Second query took 952ms
Third query took 904ms
Attached source for tests, it does not include any monetdb jars, but the versions used in the testing were:
monetdb-java-lite-2.36/monetdb-jdbc-new-2.35
Jul2017 release of normal monetDB
Hi, we have been testing MonetDBJavaLite as an alternative to "normal" monetDB and as expected we see a significant improvement when retrieving batch data (i.e. an entire column).
However we found some rather big perfomance drops when using queries like ("Select * from table where value between 100 and 300") in larger tables.
These are the result for 10 million rows and as we can see the normal monetdb becomes significantly faster following the first query while the Monetdbjavalite only have minor difference even when manually adding indexing to the table.
It seems like Monetdbjavalite are either missing some initial query optimizations that normal monetdb are using or are repeating them for each new query.
Test for Monet non embedded with 10'000'000 rows
Data generation took 3932ms
Copy binary took 1683ms
First query took 999ms
Second query took 78ms
Third query took 79ms
Test for Monet Embedded with 10'000'000 rows
Data generation took 4199ms
Copy binary took 2388ms
First query took 1031ms
Second query took 906ms
Third query took 967ms
Test for Monet Embedded with 10'000'000 rows (with explicitly added index)
Data generation took 3853ms
Copy binary took 2512ms
First query took 1092ms
Second query took 952ms
Third query took 904ms
Attached source for tests, it does not include any monetdb jars, but the versions used in the testing were:
monetdb-java-lite-2.36/monetdb-jdbc-new-2.35
Jul2017 release of normal monetDB
MonetDBJavaLiteTestFiles.zip
The text was updated successfully, but these errors were encountered: