(1) & (2) - Yes, a badly designed application that consumes too much memory can impact the performance of the server. One users interaction with such a badly designed application may impact the performance for all other users, on this or any other application. This will be related to a poor data model or an object calculation which is exploding the memory consumed. It is often only one object that causes the issues in an application when a user interacts with it.
(3) If the server hits it's max RAM resources then it may use paging memory if this is allowed by the OS which is very slow and will cause the server to appear to freeze. If you are using a VM be sure to check that the memory and CPU are dedicated for Qlik and not shared across the VM platform as this will also impact performance.
Thank you Richard. A Few follow up questions:
1) In the situation that page memory is used, and everything runs slow. What actions should we take? Are there processes we can kill?
2) We put all our CRM data in one app. Using Qlik desktop, up to 2 GB memory is used. I had to ask for additional memory. On qlik server, we'll have more than 50 users. Will the memory ussage increase linearly with each additional user?
1) I would disable the page file in windows so it does not do this. I found this link which explains how to do this on windows server 2012 as an example. https://www.youtube.com/watch?v=UZsfpXD0UmQ
2) a few points to consider.
a) to ensure your data model and application are behaving within reasonable bounds I would open it it QS desktop (as you have indicated) and click the different sheets whilst monitoring the performance monitor (windows).
b) A rule to follow for server sizing of RAM is to take the maximum size of the application when opened in QS Desktop, plus 10% of this size for each concurrent user.
You may have 50 individual users but it is likely that you will find a max of only 10 active at any one time. So...
2GB - Application opened in RAM
2GB - 10% of 2GB per concurrent active users (assume max 10 concurrent)
This would mean, for this application alone, you should assign 4GB of RAM (see note below)
PLEASE NOTE: you will need to allow RAM for the OS to perform, any reloads that occur (which could double the data in memory depending on your approach) and any other applications that are on this server instance so looking at a single application in isolation does not give you the full server spec required.