This topic aims to provide some basic information on performance to consider before you install Qlik Sense Enterprise on Windows. There are several different considerations to think about when planning your Qlik Sense Enterprise on Windows deployment:
- Size of deployment - small single-node, medium, or large multi-node site?
- Number of nodes in your site?
- Local or dedicated repository database?
- Local or network file share?
- Number of CPU cores required for each node?
- RAM required for each node?
We also recommend scalability testing and engaging with Qlik consulting services for larger deployments.
Capacity and performance
Qlik Sense supports up to a maximum of 12 nodes. In addition to the number of nodes, there are other factors that contribute to total capacity:
- Workload
- Hardware speed
- Network speed
For example, if the disk speed of the file share and the central node is too slow, you may expect low performance during some operations, such as importing or duplicating apps.
DMZ deployments
All nodes in a site, including nodes without an engine, require access to both the database and file share. In demilitarized zone (DMZ) deployments this may require opening additional ports, or taking an alternative approach, compared to a DMZ deployment with synchronized persistence.
Geographical deployments
The current persistence model does not support geographical deployments. For best performance we recommend that you locate all your Qlik Sense servers in the same geographic location or data center as the repository database and the file share with a network latency below 4 milliseconds.
Central node dependencies
The central node is responsible for handling a number of vital operations on your site. If the central node fails, some operations will fail to run, including:
- Manager scheduler - responsible for triggering reloads
- License distribution - allowing new users to obtain a license
- Extension objects
To reduce the dependency on the central node you can configure one or more nodes as a failover candidate. For more information, see Configuring failover for central node resiliency.