NOT AwareIM issue but WTF?

[BenHayat]

First question: When the client becomes slow, have you had them check the size of browser's memory & CPU usage via Task manager? If the session data that travels back & forth between client and server gets too large, then the browser becomes very slow to respond. So, when they logout the session data gets cleared. Ask them to report to you the browser's memory usage when it is slow and then have them logout and see if the browser's memory and CPU returns to normal. If, so then it's the browser's issue.

The second question I have (from my past experience dealing with uploading files) is that, are these people in the office, are they routinely upload large files, perhaps multiple users at the same time? This can take toll on the server memory which will shift to Virtual disk memory. I had used Telerik Upload component in ASP Webform, and we saw a huge memory usage on the server side of component that was accepting the files being streamed. Once the server's memory became used up, we could see the server Disk I/O gets to 100%, CPU to 100% and we knew OS was cresting disk virtual memory to deal with images, and at this time, it would affect every user.

You first need to find out, is it client side or server side.
Hope this helps!

[kklosson]

Good stuff and on target with what I have been pursuing. I have asked for the memory checks in the session but they are not very responsive. I also have a forum member crafting a client side script to limit the size of file uploads.

[BenHayat]

Good stuff and on target with what I have been pursuing. I have asked for the memory checks in the session but they are not very responsive. I also have a forum member crafting a client side script to limit the size of file uploads.

Let me tell you, I was deep in developing an app for people to archive PDF files and use at any time. I was over a year into this in ASP.Net and Telerik components, and I killed the project for ONE reason, even though I had come up with some work around. And that reason was, ASP server could not handle large file uploads, and what to speak of multiple users at the same time. ASP would hold all the streaming bytes in memory, before it would actually create a file on disk. Most of the times, client would get server timeout error, because server was not accepting stream anymore.
I'm not saying this is the case with Aware's server, but pointing out file uploads kill servers, and we ended up killing that project.

[ACDC]

I am just shooting from the hip here, have you looked into the ISP shaping the band width. Maybe they have an algorithm in place causing this or it could be a modem issue. I had a similar problem on a COX internet in CA where I had to periodically set the modem to restore normality (this was not with Aware though) After making a call to COX their customer service, it turned out there was some issue at play that was modem related and they proposed that I replace the modem. I wasn't convinced with this solution as sometimes the problem used to go away for 2 or three days and then return . It was definitely ISP related though

One of the problems when running mission critical business apps over a standard internet connection, you always going to get network issues that points to you because of the critical state of use. I find myself continuously at loggerheads with my customers IT dept and their ISP over issues that have got nothing to do with our services. Very often I have had to install sniffers and lan speed monitors on the clients network to prove my case so I don't lose the customer.

[BenHayat]

Kingsley, one other important info that might help.

When I ported my app to V7, the exact app that would run in V6 (aware as service), when run with default memory settings in V7 (on local machine), after a few minutes of testing, the browser's response would slow down and then eventually the service would pop a java memory heap error and server would crash. Never did that in V6. I reported to support in case of memory leak, but eventually support suggested to give more memory to aware service in the ClientStarter.bat I had to up the memory and never had a problem. In Java 32, 1024 is the max, but in 64 you can go to 4 gig.

%JDK_HOME%\bin\java –Xmx1024m -cp %CLASSPATH% com.bas.uiconfiguration.Application

Try this to see if it helps you. I'm sure one we go to 64-bit we can increase this to give more memory to Aware.

[kklosson]

The is the startAwareIM.bat file, right?

[kklosson]

This could be key. I had it set at 1024 in V6 and did not update the setting when I went to V7. Giving it a shot...

[BenHayat]

Yes Sir... You really need it in V7.

[BenHayat]

I had recently been reading up on GC (garbage collection) of Java and .Net, which both release "disposed memory by app". .Net version 1.x had a huge time delay releasing this memory. As of recent versions, .Net had become more responsive to give back memory to OS. In Java, it uses a different technique to give back memory to OS. It tries to use it internally before giving it back it back to OS, more like a cache memory. But this delay can cause problem for servers sometimes when then need memory. In our case Aware server (as service) can suffer low heap memory which is vital.
The 64-bit give us the ability to expose more memory to Aware at runtime, so even if Java GC doesn't immediately release memory, Aware has enough allocation to work with.
This was my understanding from reading different materials.

[kklosson]

The issue continues. After about 20 minutes, the app seems to slow, only in their session. They report that simple things, like clicking into a text box for typing, and the cursor is slow and unresponsive and they can't even highlight text without some sluggishness. So I just don't know what to make of it. This is all since V7. However, when I installed V7, I opted to stay with the 32 bit just to get a performance baseline. I provide AwareIM and Tomcat 1024m each. In the task manager, neither ever gets above 600m. I run as a service so the memory settings are set in the following files:

awareim\bin\awareim.conf:

# Initial Java Heap Size (in MB)
wrapper.java.initmemory=500

# Maximum Java Heap Size (in MB)
wrapper.java.maxmemory=1024

awareim\bin\tomcat.conf:

# Initial Java Heap Size (in MB)
#wrapper.java.initmemory=3

# Maximum Java Heap Size (in MB)
#wrapper.java.maxmemory=64


awareim\tomcat\conf\wrapper.conf:

# Initial Java Heap Size (in MB)
#wrapper.java.initmemory=500

# Maximum Java Heap Size (in MB)
#wrapper.java.maxmemory=1024

I'm a little unclear on which of these files are actually necessary so if you can shed light on that please do. All other recommendations are welcome.

[BenHayat]

I have run into the exact thing you're talking about with this site
https://directnic.com/

After a few search it starts getting slower and slower to the point you can't even type anymore. The only solution is to hit F5 and will snap back again.
While getting slow, when I look at task manager, I see FF memory and CPU is increasing more and more. So something in the client JS is just eating up more resources until refresh.

Who knows, maybe one of Kendo's widgets is eating up too much memory.

[johntalbott]

They report that simple things, like clicking into a text box for typing, and the cursor is slow and unresponsive and they can't even highlight text without some sluggishness.

I agree with Ben. Based on the above, this appears to be a client-side (JS code loaded in the browser) issue. If that is the case, server-side configuration is irrelevant.

What has to happen for the user to get back to normal performance ... close the browser and start over? Or can the user just refresh the browser?

[kklosson]

Yes. They log out/in and performance returns to normal.

[BenHayat]

What has to happen for the user to get back to normal performance ... close the browser and start over? Or can the user just refresh the browser?

I'm pretty sure a refresh (F5) will do. I tested it today. Closing browser will clear more memory than refresh.
That was in FireFox

I also run into the same issue with SquareSpace Web editor in Chrome. I have to press F5 routinely or it gets slow.

Tell user, as soon as it gets a bit slow press F5 and they don't need to logout.

[johntalbott]

Right ... I was wonder if refreshing would be enough.

Is this happening in all browser flavors and/or versions?

Here is an interesting blog post regarding a memory leak in the Kendo grid. This guy was using an older version of the library, so this could be fixed by now. But possibly not.
http://scottsjewels.blogspot.com/2013/0 ... -grid.html

Here is a very recent article that in the Kendo UI Update section mentions a fix going in for an IE specific memory leak
https://dzone.com/articles/tips-for-bec ... script-dev

I have no idea if these specific memory leaks are related to your issue, but they are things to consider.