Hi All,

Anybody have an educated guess on whether a Type 4 Pure Java driver will
ever be able to be as fast as a Type 3 driver communicating with binary
middleware on the server?

I know the Java VM's are getting faster, and it would seem that adding the
middleware layer would introduce inefficiencies, and optimizations can
almost certainly get squeezed out of the driver.

I think it is kind of a milestone that you guys have progressed the project
to the point that you are becomming more concerned with performance than
bugs or features.

Congratulations, and thanks!

Rick

----- Original Message -----
From: "Carsten Schäfer" <ca_schaefer@...>
To: <Firebird-Java@yahoogroups.com>
Sent: Wednesday, October 16, 2002 10:58 AM
Subject: Re: [Firebird-Java] Driver Speed Tests


When i look at your result with the new tests it is not only a problem with
selects,
inserts are also a problem (190% & 198% of Interclient perfomance)
What is the difference between your top tests and below printed results from
October ?

gruß
Carsten
----- Original Message -----
From: Blas Rodriguez Somoza
To: Firebird-Java@yahoogroups.com
Sent: Wednesday, October 16, 2002 4:37 PM
Subject: Re: [Firebird-Java] Driver Speed Tests


Hello

I try with your test and the conclusion is that there is a serious
performance problem in selects, but it is not related with character sets.

I create a new test from the one you send removing character set and
run the tests in Interclient and our driver. The results are

Interclient
Insert with character set = 11500
Insert without char set = 11984
Select with char set = 2156
Select without char set = 1703

Firebird Driver
Insert with character set = 22834 (198% of IC)
Insert without char set = 22812 (190% of IC)
Select with char set = 3813 (177% of IC)
Select without char set = 3188 (187% of IC)

As you can see, the problem is not in the character set use.

I compare the results of the my June tests with the actual ones. The
machine is not the same, so the absolute value is not meaningful. Also the
June tests use JDK1.3.1 and now use JDK1.4.1

Interclient (June)
Insert with Autocommit = 8510
Select with Autocommit = 3052
Insert with transaction = 6650
Select with transaction = 3990

Firebird (June)
Insert with Autocommit = 14307 (168% of IC)
Select with Autocommit = 3056 (100% of IC)
Insert with transaction = 5522 (83% of IC)
Select with transaction = 3819 (95% of IC)

Interclient (October)
Insert with Autocommit = 3043
Select with Autocommit = 734
Insert with transaction = 1484
Select with transaction = 735

Firebird (October)
Insert with Autocommit = 4918 (162% of IC) better 6%
Select with Autocommit = 1375 (187% of IC) worse 87%
Insert with transaction = 1441 (97% of IC) worse 14%
Select with transaction = 1168 (159% of IC) worse 64%

After all those numbers, it seems that there is a performance problem
in the Selects due to some code changed since June.

I'll try to locate the problem, but I don't have too much idle time
now, I hope others can help.

Regards
Blas Rodriguez Somoza

----- Original Message -----
From: Carsten Schäfer
To: Firebird-Java@yahoogroups.com
Sent: Wednesday, October 16, 2002 10:38 AM
Subject: Re: [Firebird-Java] Driver Speed Tests


Hi,
the tests was done with Win2000 SP3 on Intel 600Mhz, Sun JDK1.3.1_05.
I can give you the exact times later.

gruß
Carsten
----- Original Message -----
From: Blas Rodriguez Somoza
To: Firebird-Java@yahoogroups.com
Sent: Wednesday, October 16, 2002 10:07 AM
Subject: Re: [Firebird-Java] Driver Speed Tests


Hello

Thanks for the code, I'll try to investigate that this week.

There are some information that can help me when evaluating the
problem.

1.- Which JDK are you using?. Because our driver is pure Java,
with newer JDK's the performance enhancement is greater than in Interclient.

2.- Can you give me the test time for Interclient and Jaybird?.

Regards
Blas Rodriguez Somoza
----- Original Message -----
From: Carsten Schäfer
To: Java-Firebird
Sent: Monday, October 14, 2002 4:58 PM
Subject: [Firebird-Java] Driver Speed Tests


Hi
I've written a little test class to test the perfomance difference
between Jaybird and Interclient with varchars with character set WIN1252.

This is my db structure:
CREATE DATABASE 'd:\DBTEST.GDB'
USER 'sysdba' PASSWORD 'masterkey'
PAGE_SIZE 4096;
CREATE TABLE T_TEST1 (
F_TEST1 VARCHAR(250) CHARACTER SET WIN1252,
F_TEST2 VARCHAR(250) CHARACTER SET WIN1252,
F_TEST3 VARCHAR(250) CHARACTER SET WIN1252,
F_TEST4 VARCHAR(250) CHARACTER SET WIN1252,
F_TEST5 VARCHAR(250) CHARACTER SET WIN1252,
F_TEST6 VARCHAR(250) CHARACTER SET WIN1252,
F_TEST7 VARCHAR(250) CHARACTER SET WIN1252,
F_TEST8 VARCHAR(250) CHARACTER SET WIN1252,
F_TEST9 VARCHAR(250) CHARACTER SET WIN1252,
F_TEST10 VARCHAR(250) CHARACTER SET WIN1252
);

And this is my test class:

import interbase.interclient.CharacterEncodings;

import java.sql.*;
import java.util.Properties;

public class DBTEST {
public static final void dbtest(boolean interclient) {
try {
Properties dbProperties = new Properties();
dbProperties.put("user","sysdba" );
dbProperties.put("password", "masterkey");
dbProperties.put("charSet",CharacterEncodings.Cp1252);
dbProperties.put("lc_ctype", "WIN1252");
Connection con = null;
if (interclient) {
Class.forName("interbase.interclient.Driver");
con =
DriverManager.getConnection("jdbc:interbase://localhost/d:/dbtest.gdb",
dbProperties);
} else {
Class.forName("org.firebirdsql.jdbc.FBDriver");
con =
DriverManager.getConnection("jdbc:firebirdsql://localhost/d:/dbtest.gdb",
dbProperties);
}
long start = System.currentTimeMillis();
PreparedStatement insert = con.prepareStatement("insert
into t_test1 values (?,?,?,?,?,?,?,?,?,?)");
for (int i= 1;i<=20000;i++) {
insert.setString(1,
"ölkkkdsklsdaöd0590945650096945ß945ß46ßsakldsalk9034298428
45290904528420z5928dfldsf"+i);
insert.setString(2,
"iowerw8290nvdshdskljkfduiuj-xycäöüäpf3490384jhkffkhdfjgfdhfhdhfh92ß" + i);
insert.setString(3,
"dsahjdsalkasä3qpüp3410p4319p341jjludiouaidjilja" + i);
insert.setString(4, "hkdsuuzhdsauidsazhs" + i);
insert.setString(5,
"cx,mkjaä3420990320984390843kjghjghjf´ß" + i);
insert.setString(6,
"sdkhkjdsau98712098912pk<ä#dsaüaasdääöasööüööösa" + i);
insert.setString(7,
"sajkhjasiu7wq987e902ß0pjxycj9fgklgjkgfjkjkdkjldfjkgfjkfdkjfdkjkjgfkjldkjdfk
ljqei8üoeöälp" + i);
insert.setString(8,
"cx.,cxcöö0pßq0ßßq0ßew90idlkdslökiaoödiaopdoixüycü+yxcxpü+y" + i);
insert.setString(9,
"12309912031ß13820818308108091238901ß830183208129ß028128hcxkh" + i);
insert.setString(10, "cxklkxyhjsclisauuduiouduosuisa
uasiu128790132909871239891213209189" + i);
insert.executeUpdate();
}
insert.close();
long ende = System.currentTimeMillis();
System.out.println("Time for insert into
test1:"+((ende-start)) );
Statement select = con.createStatement();
ResultSet rs = select.executeQuery("select * from
t_test1");
while (rs.next()) {
String s1 = rs.getString("f_test1");
String s2 = rs.getString("f_test2");
String s3 = rs.getString("f_test3");
String s4 = rs.getString("f_test4");
String s5 = rs.getString("f_test5");
String s6 = rs.getString("f_test6");
String s7 = rs.getString("f_test7");
String s8 = rs.getString("f_test8");
String s9 = rs.getString("f_test9");
String s10 = rs.getString("f_test10");
}
rs.close();
select.close();
long ende2 = System.currentTimeMillis();
System.out.println("Time for select from test1:" +
((ende2 - ende)));
con.close();
} catch (ClassNotFoundException e) {
System.err.println("Error:" + e.getMessage());
e.printStackTrace();
} catch (SQLException e2) {
System.err.println("Error:"+e2.getMessage());
e2.printStackTrace();
}
}


public static void main(String[] args) {
DBTEST.dbtest(false);

}
}

With inserts Jaybird is about 30-35% slower than Interclient and
with Selects it's 70-75%.
Any comments ?

gruß
carsten Schäfer