Network Slow when using Enable Sort Order in Browse

[Petepall]

I have been using this new function in C6.1 and found that its very slow on network especially if the two stations have openned the Clients_Browse of my App. If the data reside on station A and i try to open a Clients_Browse on station B, its not too bad but still slow. If station A open a Clients_Browse, then all hell brake loose. Station B is very very slow in trying to show the Clients_Browse list.

Is there something available to be able to Sort on foreign key without using filters...? I beleibe this "Enable Sort Order" is using Filters.

This has been a major problem for me.

 

[ShankarJ]

Hi!

You have to disable Oplocks and write-caching for this. Check out

http://www.rescuemarketing.com/network.htm

for more info. Also, since you seem to be using a shared folder on a normal workstation and not a server, the performance would not be great. Enabling sort order will not work properly if a key is not found with the first element as the column and you need to make sure that the case (sensitive/insensitive) is correct.

Regards

 

[Petepall]

Thanks very much ShankarJ,

Very good info indeed.

I did a test this afternoon, using a third pc used as a Server more or less and the performance was a lot better.

I will check your advise and that document is pretty good.

I will get back to you with the result.

Thanks a million time.

Pall

 

[Petepall]

Hi again,

I did some modifications on 3 PC'S and there are more to do

1- Removed the index in Windows
2- Turn off Write Caching.
3- change the columm to reflect the key in use for the Browse. It did help on one PC but on the other PC, Nothing changed.

Like i said in my previous post, if the application is not running on the one that has the data, everything is fine.

I did not understand what you said about the following...

"make sure that the case (sensitive/insensitive) is correct"

To what are you referring to...?


Thanks again. Pall

 

[ShankarJ]

Hi Pall,

If you have a case-insensitive key and you are using a Column override for the Column Sort header, make sure you mark it as case-insensitive as well so that it generates the sort order as +UPPER(1stColumninKey),... which will allow the View engine to choose the key and your performance will be good. Otherwise, it will create an Ah-Hoc index and the performance on a network will be bad.

Regards

 

[Petepall]

Hi ShankarJ,

Thanks for the clarification and your help.

To summarize this problem, i did some more test today and if i understand it correctly, The problem seems to happen only on the browse that use "Enable Sort Order". Any other browse without this feature works great.

Just to make things clear...

1- All 3 stations run the same program from each station.

2- only one station has the Data and lets call this station, Station-1.

3- The other 2 stations will have no problem getting Data from Station-1 when using the Browse with the "Enable Sort Order" as long as Station-1 did not run the application and
start any Browse.

4- As soon as the Station-1 open-up a browse, the other stations will be very slow in aquiring data.

5- If on Station-1 we close any Browse and leave the application running only, again everything runs smooth.


6- All 3 stations will work just great if they dont use the Browse with the "Enable Sort Order".


I presume that if i had Network Problems, it would show up with any Browse that do not have that "Enable Sort Order" Option.


I wish Clarion could come up with something that could enable us to use some kind of Sort on Foreign Key without using filters.


PetePall.

 

[Petepall]

Hi ShankarJ

I forgot that last year you did give me a good idea which i will implement very soon to try it out.

. Store the first ten non-space characters of the Client Name in INSCRIPTION and define a key on that column + ClientID and use that key for sorting.¨

I beleived it will do the trick for me.
I will let you know soon.

PetePall

 

[Petepall]

Hi ShankarJ,

Well, i did all the things that you have mentionned and i dont see any improvment.

One thing i noticed, when i open a browse with the "Enable Sort Order", there is a lot of activity going on the network. If i open a browse without this "Enable sort Order", There is hardly no activity on the network and the browse comes up right away.

There must be some kind of operation going on and i cant put my finger on it.


Petepall.

 

[ShankarJ]

Hi Petepall,

Have you selected a KEY for the Browse i.e. in the Table schematic you press Edit on the Primary table of the Browse and choose a KEY (sort order)?

Regards

 

[Petepall]

Hi ShankarJ,

To your question regarding a Key for the browse...
That is what i use for sure.

To make a story short, changing "Filed Page" to "Page Filed" made quite a difference. No more delay in accessing the files.

Thanks for your help, i did learn a few things concerning what not to do with browse and Foreign Key Sort, NTFS and so on.

In the near future, i will learn a bit more about SQL and probably migrate to that.

Agin... Thanks.

PetePall

 

[ShankarJ]

Hi PetePall,

So you had it "File-loaded" which means it loads ALL the filtered records to the browse queue whereas "Page-Loaded" means that it just load one page of records to the browse queue.

Even with SQL, you need to choose "File-loaded" ONLY if the record set is relatively small. For example, a child browse showing Invoice lines can be File-loaded as it is a small record set. The advantage of File-loaded is that you can do things like re-calculating Invoice totals from the Browse queue instead of reading the records again.

Also, if you use Browse Totaling, it will behave as "File-loaded" even if you choose "Page-loaded".

Regards

 

[Petepall]

Hi ShankarJ,

That's right, it was file-loaded.
Things are better but in the future, i will have the same problem when the data file start getting bigger.

Thanks very much for your time and help.
It was appreciated.

PetePall