Clarity

Hi,  The clarity speed is very slow in our environment.  Can any one of you share with me any modifications  i need to do in order to increase the speed of Clarity?  Thanks.  

Where exactly is the speed low?How long does it take your login page to come up?How log to login?How long the open the project list?How long to open scheduled jobs page(s).  Which version?How many users? count by categgory?What db? version and fix or service packs?How are your time slices set up and jobs scheduled?  What daily and weekly maintenance you perform on Clarity and on your db?  What hardware you have number of servers, clustered or not, number of CPU's and memoryOS version and service packsjava vm configuration  Furhter information would be needed   on your Clarity usage which functionalities and how?System, financial and PM setttings and other configuration aspects.  Sorry to say there is not just one thing that can slow down the system nor do I consider there to be just 10 top best practises which will cure everything.  Martti K.    

Hello Martti,  Thanks for your reply.  I can say that when we open both login page and  projects list page , clarity speed is very slow (max. 5 mins).  Java version is : 1.6DB : OracleClarity version : 8.1.3.4545  But i need to find the details to answer your remaining questions.  I will come back to you.  Thanks.

A good response time to get the login page is 5 seconds and another 5 seconds to login.If it is more than that for all users (with much rights or just with timeentry rights) it is too much.At 20 seconds you need to do something or the users will get to to you.  Go remotely to the server and compare the times there.How many instances you have on the app server(s)? Non single sign and single sign.How much memory is used. If it is a Windows 32 box you can see from task manager.Win 32 can have max 4 Gigs of memory and javavm can use something like 1.3 Gig so you can set you memory at 1024 or 1536.It can barely take two instances.  Other OS systems can use more memory, but their javavm cannot necesarily use more, possibly some more.  Ensure also that there is adquate disk space on app and db servers.Is there anything in the app-niku log that helps?Is there anything in the access log which tells who is getting longest reponse times? If yes what are they doing?  Martti K.

Hi,  Some general points for system performance.  1. It is good if You have an Good Box with some high end configuration in which the App,DB and Reports       resides in seperate boxes.

2. Check the network speed as it is the culprit in many occasions.The above network & Box config is not in our hands but we   can co-ordinate with the network team and regarding the Box is less powerful,try to get an High end Machine.

3. You have an small section called "performance" under system administration in the 8.1 / 12 Installation guide.CA has provided some suggestions,can have a look at it and then discuss with your DBA.

4. JVM - It all depends upon how much memory is occupied.Have you checked the memory usuage during
 app & db peak times.JVM plays an very important factor in performance de-gradation.

Pls check this excellent post provided by Josh (CA).

http://caforums.ca.com/ca/board/message?board.id=CAClarityGeneralDiscussion&message.id=4479&query.id=125775#M4479

5. Set for all system / custom portlets / object views that the data will not be shown until it is filtered.

6.Stop / start the NIKU services daily during night-non peak hours.

7. You need to find the tables that are occupying the highest number of data and in that which tables are frequently
called and also ensure that tables are analyzed regularly.For Oracle Table Analyze Job. (For oracle this should be scheduled weekly).

8. Access Rights - It is recommended that all users’ rights are restricted so that they only have access to the
areas of the system which they need.Clean up:-   Remove all the Global and Instance, OBS, Participants access rights for all Inactive users.

9.Jobs:-   Run these Jobs on regular basis for data cleanup activities.But go through the (Job description / understand) in Admin Guide and then schedule it.

1. Clean User session Jobs
2. Remove Job and report Logs.
3. Delete Process Instance
4. Delete Investments.
5. Purge * Jobs.

10. Remove the system caches on daily basis.(Flush).

11. Shortcut links -->Provided some shortcut links, so the pages accessed by users will be short, the hits will be low that   will help the application performance.
   
12.Check the application performance Monitoring (NSA-->Can use Performance Tracking) & DB Health Monitoring (DBA's can help) frequently
and identify the problematic areas.

13. Pls properly configure the timeslices.Pls check this KB Article -->EC435563).

14. Check the Reports by running them (Identify Bad sql queries,Tune them.)

15. Process Engine -->Dedicated Service for the Process Engine.Pls check this KB Article -->TEC484783.  16. Datamart Job which is an Performance killer Job.If you use try to run in an weekend or if daily then Non Peak Hours.ANd also check Investment Allocation Job which should be run during Non paek Hours.So schedule the Jobs Meaningfully.    17.If you have an cluster environment (with many app servers,db,bg)you can use an Load balancer to        distribute work load.   cheers,  sundar

   a couple more simple checks:- run Oracle Table Analyze job once or a couple times a week- make sure sql trace did not get left on on some users (select * from cmn_sec_users where sqltrace_active is not null)- debug is not turned on when not needed in NSA    

When I ran the above query in our database I found 1 affected row
that has a particular resource name. Does that mean anything?.
it has sqltrace_value as 0 .

You could look at
TEC435531 Using Clarity SQLTrace
TEC439280 How to read SQL Trace Log Timings
TEC513073 Using Clarity SQLTrace - Explanation of Logging Levels

The purpose of the query is to see if there is SQL trace left on by accident.
The implied proper way is that if the trace is not on on purpose the value should be null not 0

Martti K.

The other way to look at it is to see if there are traces in ..\Clarity\logs\SQL for that particular user.

Martti K.

You can also check if there are any heavy duty portlets on the Overview page of the user.

Usually, this can also cause slow performance for a user

Regards
NJ

How would you relate heavy duty portlets to trace being on or not for a user?

Martti K.

nothing to do with trace ... if there are heavy duty portlets on the overview screen, it can cause slow logons as well.

NJ

hi all,

for our case, slow response occurs during project list loading and project save. what would be the possible cause and how to overcome?

thanks in advance

dott wrote:

slow response occurs during project list loading and project save. what would be the possible cause

Problems in your database - get your DBAs to investigate.

e.g. if the INV_INVESTMENTS table is poorly organised (at a database level) this would be the symptom.

Your DBAs need to do a health-check on your database to start off any performance analysis ; a slow database will really kill Clarity performance.

Do I understand it right the problem is

for our case, slow response occurs during project list loading and project save. what would be the possible cause and how to overcome?

an in general the performance is not a thing to grind your teeth about?

If that is the case
How slow is slow response?
How many projects in your system?
Which Clarity and db version to see if they have a "reputation"
If you clock the response time to open the project list for a user with all global rights to projects and a regular PM with rights to a portion of the projects. Is there a difference?
How many custom attributes and subobjets for the project object?
When you say save, do you mean save from Clarity. If you mean save from a scheduler, then which one?
How many tasks, resources and assignments is that?
Refer to Rob's benchmark
12.1.2 & Open Workbench 1.2.1 speed test results
56647912


Martti K.

two items that i've found that can affect project list load times are:

1. check number of columns that are listed on the project list view - reduce if possible for this user, as an isolation test (eg start, with just three columns, code, name, manager, see if the performance is much better, then progressively add in more to see which ones are causing performance issue.) This is looking at it from a 'columns' perspective.

2. Check the way that the user is granted access to the projects. if it is many many obs-based, complex security granting, this can cause slowness. as can instance by instance granting of security... as can many many partitions. this is looking at it from a 'rows' perspective...

This is an interesting topic and there could be many suggestions as we have seen on this thread,

My 2 cents would be:

1> If there are columns using aggregates in the List view, then I would get them out if I could, after consulting the business users.
2> Admin tool -> Objects >Project -> Views -> Project List [Options] -> Filter and choose the radio button of "Do not show results until I filter"
3> Get the SQL trace of the specific action that appears very slow in response and analyze the trace log file.

HTH

Mayank Joshi