(12.03-en) Performance Optimizations in IGEL UMS

Data Sizing 

• The number of registered firmware versions has the largest impact on the size of the database.  
  (Listed in UMS Console under Misc > Firmware Statistics)
• The number of devices or profiles has a minor impact.
• Average size per... 
  • Firmware configuration: ~15 MB 
  • Profile (depends on the number of active parameters): ~100 kB 
  • Device: ~100 kB
• Reserve 500 MB up to 1 GB for database transaction logs of excessive database calls like Remove unused Firmware. Please note that the usage depends on the database system used.    

Latencies 

If you are struggling with long-distance connections and high latency, please consider the following recommendations: 

• Minimize latency between...  
  • Database <-> UMS Server: <= 20 ms
  • Several UMS Servers: <= 50 ms
  • Load balancer <-> UMS Server: <= 50 ms 

• High latency between the database and the UMS Server has a huge impact on the performance. The communication between the device and the UMS Console will slow down, the UMS Console itself will become lazy.
• High latency between the device and the UMS Server has little impact on overall performance. 

Performance Optimizations 

• UMS logs:  
  Use administrative tasks to automatically clean up logs (logging data, job execution data, execution data of administrative tasks, process events, asset information history) or remove old UMS log files (/rmguiserver/logs)  when storage space runs out.
• Firmware:  
  Remove unused firmware regularly.
• Embedded database only:  
  • Optimize database regularly (UMS Administrator application, e.g. once a month)
  • Check for free storage space and expand the storage size if necessary (keep at least 1 GB free at all times)
• Number of devices:  
  • If the device count is high (>10k) and overall performance is low, increase UMS Server and UMS Console memory. See (12.03-en) How to Configure Java Heap Size for the UMS Server and (12.03-en) How to Configure Java Heap Size for the UMS Console.
  • Avoid too many devices (>5k) in one folder.
• Assignments:  
  Keep the number of assignments per device (direct and indirect) at a low level (<25).
• Administrative tasks and jobs:
  The more administrative tasks and jobs are created, the more heap is "eaten up", so it may be necessary to increase UMS Server memory. See (12.03-en) How to Configure Java Heap Size for the UMS Server.
• Default directory rules:  
  Do not use default directory rules with the Apply rule when device boots option unless they are required.
• Concurrent device requests: 
  If you are experiencing problems with many concurrent device requests (delays in configuration deployment or logging on to the device), open the UMS Console and use the options under UMS Administration > Global Configuration > Device Network Settings > Device Requests  (thread and queue size) to control the throughput of the device requests. Contact support for recommendations. 

Limitations: UMS HA 

• Device actions that are manually triggered in the UMS Console are performed by one UMS Server (the one the UMS Console is currently connected to); there is no load balancing for these actions.