Inconsistent Load Times Between Qlik Sense Engines with Load Balancing

[Alvin_wei]

Hi @marcus_sommer , 

I'm running a Qlik Sense application with four engines configured for load balancing. When I create a new session, the first binary loads the script from a remote server A, and then the data files are loaded using the LOAD * FROM [LIB://xxxxxx] statement, also from server A.

Even though the engines have the same CPU and memory specifications, I'm noticing inconsistent load times between the engines. Some engines are significantly faster than others when loading the same data. I'm trying to understand the potential reasons for this and how I can debug the issue.

[marcus_sommer]

At first make sure that all loads you are looking for are identically - a binary load isn't the same as loading from qvd's or other file-types. Also that there are no differences to where-clauses, data-base drivers and data-interpretations (interpretation-variables, geo-location from the OS + Qlik installation and similar stuff) because it could have an impact to the number of loaded/stored records.

The next might be to look for the OS + Qlik releases which should be the same and investigating the network because the distance between the servers and the storage as well as the priority-control could be different (don't forget proxy- and balancer-clusters).

Another aspect are the VM configurations. I assume that all nodes as well as all other parts of the environment are VM and looking from the surface they may look equally but the underlying hosts may reside on different hardware and may sharing the resources with n other tools or not and/or having a lot of other different settings.