started at 9:30 so as to verify the results..interestingly found that addentry was working correctly. Wanted to test if we can generalize to any sequence diagram...
I will try to prove theoretically if we can check this out
Monday, September 29, 2008
Tuesday, September 23, 2008
work on sept 23
Plan
1) compare response time given by autoperf and httperf with no of users =1 for both view entry and write entry
2) check if we can measure the response time for Login transactions, problem was it wasn't calling week.php, if we do login
3) Finally do measure response time for viewEntrybyWeek transaction and write Entry transaction, login using httperf and autoperf.(p 1hour)
4) check out how service demand is being calculated for database server in autoperf. i.e why service demand went on increased (p3 hours )
5)check out tracers for both php and autoperf(p1 hour)
*
1) There was a problem that httperf output had 4XX=20. the problem was with input scripts written by me.
Httperf results with 1 req/sec is 108.8 ms
autoperf results with 1 req/s is 156 ms keeping thinktime 0.85 ms and utilization of webserver is 14 % and utilization of database is 0.97 % based on sar. but autoperf shows 14% and 0 % may be it cannot capture at this level
*autoperf results with 1 req/s is 165 at think time 1s
2) login.php , mistake done was assuming that ${y} generated in a particular transaction is constant. It is working, so we have three transactions login, viewentry, addentry on which i can measure performance.
3)I can do it so I will finish it off tonight and check results
4)
additional points:
I think, measurement of service demand at mysql from autoperf is wrong, so i am planning to measure the performance by writing some sample PHP which has all the sql queries in it. so it is similar to normal viewentry.
5)i may have to work on eclipse for autoperf, but what about tracer on autoperf-linux profiler and there are php debuggers
worked for 3 1/2 hours
started working from 9 in the evening..
Ran for login scenario on autoperf
1) compare response time given by autoperf and httperf with no of users =1 for both view entry and write entry
2) check if we can measure the response time for Login transactions, problem was it wasn't calling week.php, if we do login
3) Finally do measure response time for viewEntrybyWeek transaction and write Entry transaction, login using httperf and autoperf.(p 1hour)
4) check out how service demand is being calculated for database server in autoperf. i.e why service demand went on increased (p3 hours )
5)check out tracers for both php and autoperf(p1 hour)
*
1) There was a problem that httperf output had 4XX=20. the problem was with input scripts written by me.
Httperf results with 1 req/sec is 108.8 ms
autoperf results with 1 req/s is 156 ms keeping thinktime 0.85 ms and utilization of webserver is 14 % and utilization of database is 0.97 % based on sar. but autoperf shows 14% and 0 % may be it cannot capture at this level
*autoperf results with 1 req/s is 165 at think time 1s
2) login.php , mistake done was assuming that ${y} generated in a particular transaction is constant. It is working, so we have three transactions login, viewentry, addentry on which i can measure performance.
3)I can do it so I will finish it off tonight and check results
4)
additional points:
I think, measurement of service demand at mysql from autoperf is wrong, so i am planning to measure the performance by writing some sample PHP which has all the sql queries in it. so it is similar to normal viewentry.
5)i may have to work on eclipse for autoperf, but what about tracer on autoperf-linux profiler and there are php debuggers
worked for 3 1/2 hours
started working from 9 in the evening..
Ran for login scenario on autoperf
Friday, September 12, 2008
Saturday
1) Started at 10:30
Things done while moving Webcal, just copy webcal(i.e all PHP source files) folder to webserver directory.
Check out System administrators guide which will be poped up when you open webcal in browser. for e.g
http://10.105.41.26/webcal/docs/WebCalendar-SysAdmin.html contains system administration guidelines.
2) Change permission of includes:
chmod 777 includes
3) install MYSQL on some server, it is easy you can do it, login using root and then type following commands
some other useful commands
mysql> show tables;
+--------------------------+
| Tables_in_intranet |
+--------------------------+
| webcal_asst |
| webcal_categories |
| webcal_config |
| webcal_entry |
| webcal_entry_ext_user |
| webcal_entry_log |
| webcal_entry_repeats |
| webcal_entry_repeats_not |
| webcal_entry_user |
| webcal_group |
| webcal_group_user |
| webcal_import |
| webcal_import_data |
| webcal_nonuser_cals |
| webcal_reminder_log |
| webcal_report |
| webcal_report_template |
| webcal_site_extras |
| webcal_user |
| webcal_user_layers |
| webcal_user_pref |
| webcal_view |
| webcal_view_user |
+--------------------------+
23 rows in set (0.00 sec)
Once Database installation is done, change settings in /etc/mysql/my.cnf change bind address to address of ip address on which mysql is installed.
Finished this work by 12:10.
Setting up for performance testing:
moved slaves to database and web server, master is at jacobson.
checked if frequency scaling governors enables on those PCs using
cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor.
changed it to performance so that it operates at high frequency using:
echo performance > /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor
Finished this at 12:25
Check out if the server and client has sysstat package for autoperf to work so install it using sudo apt-get install sysstat
service time at Database server was 7ms for initial run, ran intensive between 1 to 15 users on webcal application before leaving for lunch at 12:45.
Resumed at 2:45 , there was problem with the run, i had to run once more and in mean time i will try to find some PHP application till 3:45.
Resumed work at 9:00 PM devecot imap is installed ( i couldnot install other imap servers, courirer,uw,cyrus), tried to find out the problem with SMTP( finallyy installed sendmail, i haven't installed SMTP had some problem) sendmail is also MTA.
With webcalendar problem of increase in database still persists i.e for initial iteration service demand was 8 ms and as you increased number of users service demand went to 160ms, and
squirrel mail installation guide is at
https://help.ubuntu.com/community/Squirrelmail
Debain installation link
http://www.brennan.id.au
link has a niceway to connect to MYSQL from PHP.

Things done while moving Webcal, just copy webcal(i.e all PHP source files) folder to webserver directory.
Check out System administrators guide which will be poped up when you open webcal in browser. for e.g
http://10.105.41.26/webcal/docs/WebCalendar-SysAdmin.html contains system administration guidelines.
2) Change permission of includes:
chmod 777 includes
3) install MYSQL on some server, it is easy you can do it, login using root and then type following commands
some other useful commands
once everything is done output of show tables should be ;SHOW DATABASES;create database intranet; create user webcalendar;
Show Tables;
GRANT ALL PRIVILEGES ON *.* TO webcalendar@kahn.cse.iitb.ac.in IDENTIFIED BY 'webcal01' WITH GRANT OPTION
use intranet; //then
copy table-mysql.sql file and copy it on console showing mysql>. check if it is successful, if it is
successful you should see the following.
MYSQL reference:
http://dev.mysql.com/doc/refman/5.0/en/creating-tables.html
mysql> show tables;
+--------------------------+
| Tables_in_intranet |
+--------------------------+
| webcal_asst |
| webcal_categories |
| webcal_config |
| webcal_entry |
| webcal_entry_ext_user |
| webcal_entry_log |
| webcal_entry_repeats |
| webcal_entry_repeats_not |
| webcal_entry_user |
| webcal_group |
| webcal_group_user |
| webcal_import |
| webcal_import_data |
| webcal_nonuser_cals |
| webcal_reminder_log |
| webcal_report |
| webcal_report_template |
| webcal_site_extras |
| webcal_user |
| webcal_user_layers |
| webcal_user_pref |
| webcal_view |
| webcal_view_user |
+--------------------------+
23 rows in set (0.00 sec)
Once Database installation is done, change settings in /etc/mysql/my.cnf change bind address to address of ip address on which mysql is installed.
Finished this work by 12:10.
Setting up for performance testing:
moved slaves to database and web server, master is at jacobson.
checked if frequency scaling governors enables on those PCs using
cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor.
changed it to performance so that it operates at high frequency using:
echo performance > /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor
Finished this at 12:25
Check out if the server and client has sysstat package for autoperf to work so install it using sudo apt-get install sysstat
service time at Database server was 7ms for initial run, ran intensive between 1 to 15 users on webcal application before leaving for lunch at 12:45.
Resumed at 2:45 , there was problem with the run, i had to run once more and in mean time i will try to find some PHP application till 3:45.
Resumed work at 9:00 PM devecot imap is installed ( i couldnot install other imap servers, courirer,uw,cyrus), tried to find out the problem with SMTP( finallyy installed sendmail, i haven't installed SMTP had some problem) sendmail is also MTA.
With webcalendar problem of increase in database still persists i.e for initial iteration service demand was 8 ms and as you increased number of users service demand went to 160ms, and
squirrel mail installation guide is at
https://help.ubuntu.com/community/Squirrelmail
Debain installation link
http://www.brennan.id.au
link has a niceway to connect to MYSQL from PHP.

Things useful while using webcalendar
problems with autoperf:
It cannot handle when webserver returns error pages and it would be better if it waits between two kind of load generations.
http://webcalendar.cvs.sourceforge.net/*checkout*/webcalendar/webcalendar/docs/WebCalendar-SysAdmin.html
initial guide on HTTPERF
http://shahidz.com/http-performance-testing-with-httperf-autobench-openload/
How to set no of threads in apache server?
apache2 conf is at /etc/apache2/ and httpd conf is at /usr/local/apache2/bin/
I don't know why i have both httpd and apache2
I have changed httpd-mpm.conf to all threads being at 10
How to set no of threads in MYSQL server?
> vim /etc/mysql/my.cnf is the place where you can change mysql configurations
SQL commands useful
> delete from webcal_entry where cal_priority=2
> create database intranet;
> create user webcalendar;
> GRANT ALL PRIVILEGES ON *.* TO webcalendar@kahn.cse.iitb.ac.in;
> GRANT ALL PRIVILEGES ON *.* TO webcalendar@kahn.cse.iitb.ac.in
> IDENTIFIED BY 'webcal01' WITH GRANT OPTION;
It cannot handle when webserver returns error pages and it would be better if it waits between two kind of load generations.
http://webcalendar.cvs.sourceforge.net/*checkout*/webcalendar/webcalendar/docs/WebCalendar-SysAdmin.html
initial guide on HTTPERF
http://shahidz.com/http-performance-testing-with-httperf-autobench-openload/
How to set no of threads in apache server?
apache2 conf is at /etc/apache2/ and httpd conf is at /usr/local/apache2/bin/
I don't know why i have both httpd and apache2
I have changed httpd-mpm.conf to all threads being at 10
How to set no of threads in MYSQL server?
> vim /etc/mysql/my.cnf is the place where you can change mysql configurations
SQL commands useful
> delete from webcal_entry where cal_priority=2
> create database intranet;
> create user webcalendar;
> GRANT ALL PRIVILEGES ON *.* TO webcalendar@kahn.cse.iitb.ac.in;
> GRANT ALL PRIVILEGES ON *.* TO webcalendar@kahn.cse.iitb.ac.in
> IDENTIFIED BY 'webcal01' WITH GRANT OPTION;
Tuesday, September 2, 2008
Change of environment will it help?
Planned to work for more than 5 hours a day on MTP, starting from 03rd september, hope this turns out to be good, I know the least thing that i want from our guide is not to be scolded. So planning to do it sincerely. This blog contains journey of my MTP day-to-day activities. Planning to update it daily at 11:30 PM. BTW I am writing this on Vinayaka-chavathi..hope it works out to be the successfully journey...
Subscribe to:
Comments (Atom)