Feature documentation: Difference between revisions

Jump to navigation Jump to search
no edit summary
No edit summary
No edit summary
Line 34: Line 34:


1. System overload and skipped packets.  
1. System overload and skipped packets.  
If packet drops appear, the packet analyzing threads might be overloaded. This can be checked in the load statistics.  
:If packet drops appear, the packet analyzing threads might be overloaded. This can be checked in the load statistics.  
The current load percentage can be clicked to get detailed information about each system component.
The current load percentage can be clicked to get detailed information about each system component. If the analyzing threads tend to have significantly more load than the DB threads, it is recommended to increase the '''DB thread factor''' in allocate less DB threads and use more analyzing threads.
If the analyzing threads tend to have significantly more load than the DB threads, it is recommended to
increase the “DB thread factor” in allocate less DB threads and use more analyzing threads.




2. System overload due to DB message lost.
2. System overload due to DB message lost.
If DB message lost counters appear in the detailed [[System_Info_Page|load]]  statistics (by clicking on the current load percentage), it means that the DB threads where not capable of processing the data created by the analyzing threads.
:If DB message lost counters appear in the detailed [[System_Info_Page|load]]  statistics (by clicking on the current load percentage), it means that the DB threads where not capable of processing the data created by the analyzing threads.


In this case it is beneficial to decrease the “DB thread factor” to use more DB threads and thus decrease the load per thread.
In this case it is beneficial to decrease the '''DB thread factor''' to use more DB threads and thus decrease the load per thread.
Also, if the load graph shows only sporadic peaks to 100% load, it can help to increase the message queue size to buffer temporary overload.  
Also, if the load graph shows only sporadic peaks to 100% load, it can help to increase the message queue size to buffer temporary overload.  
The “Queue multiplier” is an integer value which is used to multiply the queue size by this value. The value should be chosen carefully as a value of 10 already makes the queue ten times larger but also uses ten times more memory and can also increase the latency of the displayed measurement values by a factor of ten.
The '''Queue multiplier''' is an integer value which is used to multiply the queue size by this value. The value should be chosen carefully as a value of 10 already makes the queue ten times larger but also uses ten times more memory and can also increase the latency of the displayed measurement values by a factor of ten. If regular overload appears, a larger queue usually does not help much as it will just delay the message lost.
If regular overload appears, a larger queue usually does not help much as it will just delay the message lost.
1,775

edits

Navigation menu