VM vs Physical Server Setup for High-Reload Qlik Sense Multi-Node Environment

[durgesh22]

Hi All,

We are running Qlik Sense Enterprise on Windows (on-prem) in a multi-node environment:

• 1 Central Node
• 7 RIM Nodes → 2 Proxy Nodes + 5 Scheduler Nodes
• Maximum App Size: ~5 GB
• Reload Volume: ~10,000 reloads/day (most finishing within 1 minute)
• VM Specs: 96 GB RAM, 24-core CPU per node

We are evaluating our infrastructure options and would appreciate your advice based on experience:

1. What would be the ideal setup in this case?
2. Should we move to a multi-node Physical server setup with higher specs (256–512 GB RAM)?
3. Would a hybrid approach make sense — e.g., Central and Failover node on Physical servers, and Scheduler nodes on existing VMs?

If anyone has done similar migrations or performance comparisons between VM and Physical servers for high-reload environments, please share:

• Performance impact you observed
• Cost–benefit considerations
• Any best practices or pitfalls to avoid

Thanks in advance for your insights!

[BTIZAG_OA]

Hello there,

If you are looking for performance you can switch the Schdule nodes to Phsyical ones. Generally phsyical servers outperforms the VMS because of dedicated hardware. General approach i use is go all engine operations(Schedule/Serve HUB/ODAG...) to physical nodes for better performance & Other Qlik roles(Central/Proxy/Fileshare...) goes for VMS for more flexibility. On your scenario i do not see any engine node, only schedule nodes. First of all i would ask why you want to switch to phsyical nodes? You saying most of your jobs are completed in 1 minute so there shouldnt be a any user complaintments. Since your deployment is heavily on schedule, i would consider calculate RAM at first. if 96GB VMS can handle the ram load 512 would be a too much for physical. Keep that in mind, bigger RAM sizes is not always mean faster app loads, its only serve more apps. App Load speed depends on Memory MHZ speed(On hardware side, also app model&appsize have impact too). You need to take look at the current server's RAM usage during the busiest hours of the day. This amount of RAM will show you the minimum limit of your physical server's needs. If the largest app is 5gb on your case, you do need so much RAM since Schedule means refreshing the data and its only inital loads the app on the server(it do not calculate user session & selections cache) + they are flushed after 8hours by default if you dont change the app cache time. You can consider the increase CPU Core count in order the increase parallel taks runs if needed.

I would probably set up the deployment like this,

VMS
1-Central Node > Set as MasterScheduler
1-Central FailOver Node
2-Proxy Nodes

Phsyical
5-Scheduler Nodes > Set as WorkerScheduler + increase the MaxConcurrent Reloads to = CoreCount-2

I hope this helps, if you provide more info about your needs & purpose the switching to phsyical nodes i'd be gladly to help you
Have a nice day