Hi,

okay, I'll try:

<memory maxmem="350" /> sets the limit for the memory which is used for the decoded images,

when a image is downloaded, for example with image size 1024x1024,
then this image needs 4 MB of memory (4bytes/pixel = 1024x1024x4),

on very big multiresolution panoramas a lot of images were downloaded,
if all images were saved in memory, the system memory limited will be reached soon,
and the system will begin to swap and slow down or crash,

so when this limit is reached then the memory of the latest used images
is freed and returned to the system, when a images is needed again
it will be decoded again,

here is also a documentation of it - http://krpano.com/docu/multires/

I set the limit to 350 MB, this should be enough for the most cases


"network.downloadqueues"
- the number of simultaneous downloads
- limited by the browser (FF2=2, FF3=6, IE7=2)
- non-blocking
- high values are good for fast internet connections and fast servers

"network.decodequeues"
- the number of simultaneous image decodes
- the decoding of the images is a "blocking" operation, that means it's not possible to move in the pano while decoding (a flash limitation!)
- show be a low value, only if the images are very small (small tilesizes) it can be increased


everything clear?

Hi Klaus,

I am working on a pano based Intranet project where we are looking at having several thousands (3-4000) users accessing the application/site at the same time.

The problem will not be so much with the panoramas I think, as each individual panorama is loaded once and then cached locally, and loaded into RAM from the cache after the first download from the server. Can you confirm this?

The big issue will rather be with the handling of streaming flash videos inside some of the panoramas. If you have a broadband type of video streaming at 300 kpbs then 1000 simultaneous users will require 300 megabits per second of bandwidth!

I am trying to come up with answers to our client for questions on bandwidth usage per user as well as issues on scaling the site / application to the required amount of users and last, but not least coming up with a minimum hardware configuration for using such an application: CPU speed, memory, OS, Flash version and IE version. So far I have used the Adobes System Requirement stats for Flash 10, but I have a feeling that 128Mb RAM is too little when dealing with 8k panoramas and streaming video?
http://www.adobe.com/products/flashplayer/systemreqs/

What we are planning to do is to set up several virtual machines on a Vmware server with different RAM configurations and CPU speeds to test the performance of the panorama application, so I would like to be able to measure what happens for instance when there is insufficient memory available for krpano.

Is there a way to show the memory usage in krpano in realtime ?

Cheers,
Kim

Hi,

about the downloading and caching - this is completely handled by the overlaying browser,
it's like opening the files normally in the browser, if the browser cache is large enough also
video files can be completely stored in the cache,

Zitat von »kimblarsen«

Is there a way to show the memory usage in krpano in realtime ?

partially - e.g. the options plugin shows the current memory usage of the current
downloaded images and of the decoded images,

here is a screenshot:


but flash uses also memory for many other things...

Thanks for your replies Klaus and Graydon.

Graydon - I don't think putting the flash video files on a server outside of the Intranet will make much difference. The data will still have to travel on/onto the Intranet network. Controlling the number of simultaneous video streams is the key here.

I just got into contact with some network/server consultants that will help me define a framework. It seems you can set up an Adobe Media Server with something called streaming collections where the server is sending say 20 video streams every second and then the client will hook onto the nearest available free video stream provided the intranet supports multicast. This is all above my head, but I am trying to wrap my head around it.

If we get a go to create this application it will most likely take a while before we can publish a version that is accessible to people outside that company, but hopefully I can publish some tech demos with related content for the benefit of the krpano and pano community at large. In any case I will try to be as open about the project as I possibly can.

Cheers,
Kim