jBASE Performance Tuning

[Shiva Shankar]

Hi,

I am currently working on jBASE 4.1.5.17
Could anybody throw some light on the Environmental Variables to be set once jBASE is installed on a server to achieve the best performance? To be more clear, please suggest the General Parameters that needs to be addressed after Installing jBASE.

I under stand that the question put is very general, but there are few processes which are taking thrice the time than it took on Universe.

Thanks and Regards
Shiva

[Jim Idle]

Well, you need to tell us the operating system, what processes are taking longer, whether this is the same hardware, same filesystem config and as much else as you can think of, even if you don't think it is relevant. However:

1) Did you resize the files with jrf?
2) Have you tuned the operating system in general?

Please take a moment to read the posting guidelines, linked at the bottom every post to the group - this will help us to help you.

Jim

[Shiva Shankar]

Hi Jim,

OS: Compaq tru64
jBASE: 4.1.5.17
T24: R06.008

Major Accounting files were Resized: STMT.ENTRY / CATEG.ENTRY / RE.CONSOL.SPEC.ENTRY.

The above accounting files are distributed.

I do not have much knowledge on the server side tuning and settings.
While running the COB, the output of WHERE )V, mw42 and SHOW-ITEM-LOCKS was being monitored constantly. There was no deadlock situation or locking issue. COB was run using 3 agents and i did not observe or encounter a situation where the agents had a lock on the same record.

Do SELECT's on jBASE take more time than on UNIVERSE?

Best Regards
Shiva

[Jim Idle]

On Tue, 2008-08-12 at 12:45 +0530, Shiva Shankar wrote:

Hi Jim,

OS: Compaq tru64

Tru 64 5 was the best UNIX ever released and of course therefore subsequently wiped out by HP to my eternal regret.

jBASE: 4.1.5.17
T24: R06.008

Major Accounting files were Resized: STMT.ENTRY / CATEG.ENTRY / RE.CONSOL.SPEC.ENTRY.

The above accounting files are distributed.

OK - perhaps the partition algorithm is not effective with the modulos you have chosen for the part files. It depends on the access pattern, which I cannot predict (and perhaps you can't if this is the T24 application itself).

I do not have much knowledge on the server side tuning and settings.
While running the COB, the output of WHERE )V, mw42 and SHOW-ITEM-LOCKS was being monitored constantly. There was no deadlock situation or locking issue. COB was run using 3 agents and i did not observe or encounter a situation where the agents had a lock on the same record.

Do SELECT's on jBASE take more time than on UNIVERSE?

Not normally on 4.1. However, all SELECTs will benefit from the creation of indexes.

One thing that most people forget to do is to defrag their UNIX file system, which sometimes helps depending on how the files get built up (they are hole filled at creation time which usually makes them contiguous). Most filesystems can defrag online (except for ext3 on Linux - switch to XFS if you are using ext3 you Linux guys) and that is worth a shot. You can also achieve a form of defrag on individual files by dd'ing them to copies and deleting the original or using jrf while nobody else is using the system. The command is:

man defragment
or
man defragcron

To see how to run that.

The other thing is that some constructs will be faster than UniVerse and likely some slower. However, there is almost always a way to use a faster jBASE method once you know what you are looking for of course. So, a good place to start is to pin down just exactly what is going slower, as in a specific SELECT statement with the dictionary elements it is using.

Finally, for the moment, make sure that you have allocated enough memory for the filesystem buffers. I can't remember if Tru64 auto tunes this or whether you have to tell it how much it can use. Start with:

man -k sys_attr

Which will give you good pointers.

Finally, finally - of course makes sure that you are comparing like for like and are running with the same memory configuration, same filesystem configuration and so on. Also, do TEMENOS not offer to help you tune the system after an upgrade?

I am told, but do not know, that the following is a good resource:

http://www.elsevier.com/wps/find/bookdescription.cws_home/677529/description#description

As is this:

http://books.google.com/books?id=Lmkl4otf9fkC&pg=PA395&lpg=PA395&dq=Tru64+tuning+guide&source=web&ots=Lf9TvfdgO0&sig=WNyLeVPAVRQqWdyxPq2RDBssIh0&hl=en&sa=X&oi=book_result&resnum=5&ct=result#PPA402,M1

Jim

[Tec Murho]

Do you still have same problem
Is this happens more time when u launch SELECT command
If so can u check about your file system
if u have not well adjusted ur FS u will have problem of delaying.
If for example u have one HDD and its mirror in a file system the process will take a long to select from it. you must be carefull and just see your server house of how u can well adjust ur FS
Hope this will help you
Best regards

[Shiva Shankar]

There has been a few issues with the server settings.

How are the settings of the ulimit -a going to affect your performance?
 

Also with respect to the CACHE memory and limit set, are they related to the system performance?
 

--
Thanks and Regards
Shiva Shankaran.S

[acleda]

Please check your I/O disk , sometime the performance is caused by the
I/O speed.... your disk array is RAID-5 or RAID-10?


Apsara.

On Aug 21, 3:39 pm, "Shiva Shankar" <aqua.shanka...@gmail.com> wrote:
> There has been a few issues with the server settings.
> How are the settings of the ulimit -a going to affect your performance?
>
> Also with respect to the CACHE memory and limit set, are they related to the
> system performance?
>

> On 8/18/08, Tec Murho <tecmu...@gmail.com> wrote:
>
>
>
>
>
>
>
> > Do you still have same problem
> > Is this happens more time when u launch SELECT command
> > If so can u check about your file system
> > if u have not well adjusted ur FS u will have problem of delaying.
> > If for example u have one HDD and its mirror in a file system the process
> > will take a long to select from it. you must be carefull and just see your
> > server house of how u can well adjust ur FS
> > Hope this will help you
> > Best regards
>

> >> deleting the original or using jrf *while nobody else is using the system
> >> *. The command is:

>
> >> man defragment
> >> or
> >> man defragcron
>
> >> To see how to run that.
>
> >> The other thing is that some constructs will be faster than UniVerse and
> >> likely some slower. However, there is almost always a way to use a faster
> >> jBASE method once you know what you are looking for of course. So, a good
> >> place to start is to pin down just exactly what is going slower, as in a
> >> specific SELECT statement with the dictionary elements it is using.
>
> >> Finally, for the moment, make sure that you have allocated enough memory
> >> for the filesystem buffers. I can't remember if Tru64 auto tunes this or
> >> whether you have to tell it how much it can use. Start with:
>
> >> man -k sys_attr
>
> >> Which will give you good pointers.
>
> >> Finally, finally - of course makes sure that you are comparing like for
> >> like and are running with the same memory configuration, same filesystem
> >> configuration and so on. Also, do TEMENOS not offer to help you tune the
> >> system after an upgrade?
>
> >> I am told, but do not know, that the following is a good resource:
>

> >>http://www.elsevier.com/wps/find/bookdescription.cws_home/677529/desc...
>
> >> As is this:
>
> >>http://books.google.com/books?id=Lmkl4otf9fkC&pg=PA395&lpg=PA395&dq=T...

>
> >> Jim
>
> --
> Thanks and Regards

> Shiva Shankaran.S- Hide quoted text -
>
> - Show quoted text -