[FX.php List] Web traffic versus native FileMaker use

[Jonathan Schwartz]

Dale,

Thx for your input.

My online system performs perfectly for a good part of the day.  I 
have some pretty extensive logging and use one particular frequent 
operation as a baseline. It takes 0.1 second each time. This rules 
out problems with the web operation.  On the other hand, I can 
confirm that the native FileMaker system has tons of calculated 
summary fields and filtered portals...the kind that you have to sit 
and wait for....under the domain of a different developer.  So, it is 
safe to conclude that the native system bogs down under load.

That being established, the question is how much can I expect from 
the online system if the native system has these issues?  Throwing 
big iron at it (2 machine deployment) didn't solve the problem. The 
degradation is a sudden one from baseline 0.1 seconds to 30 seconds 
for a couple of minutes and then back down again.  All the while, the 
Windows Task manager practically shows a picture of the machine 
yawning.

It could be sorting. It could be a script. It could be summarizing in 
native mode....or all three. I have no control over any of these.

It looks like I'm talking myself into a new piece of hardware. ;-)

Jonathan


At 1:48 PM -0500 5/6/10, Dale Bengston wrote:
>Hi Jonathan,
>
>I'll bite. This happened to me when I put a web UI on another 
>developer's FMP client/server installation. Creation of an order 
>line item started to take longer and longer, until the point of 
>absurdity. I finally traced it to some calc fields used by the 
>client/server UI, that summarized data. (Running totals, etc.) In 
>the client/server environment, values of constants/global fields 
>made sure those line items were only summarized for the current 
>found set (i.e. the invoice you were viewing). But on the web, the 
>entire table was getting summarized. Once the number of line items 
>got in the thousands, web performance went into the tank. This 
>problem did not manifest itself during testing, when there were just 
>a few test invoices at any given time. Note that these fields were 
>not on any layouts used by the web. They were calculated on record 
>creation.
>
>My problem was based on volume of records, rather than time of day, 
>but maybe my experience will help you triangulate. Does your 
>performance get better with fewer clients logged in? What kind of 
>ongoing client/server scripts are running, even when clients are 
>idle?
>
>Hope this helps,
>Dale
>

-- 
Jonathan Schwartz
Exit 445 Group
jonathan at exit445.com
http://www.exit445.com
415-370-5011