locked
High incoming message processing time RRS feed

  • Question

  • Hello,

    I installed the Real time server monitoring for Skype server and I see that the value "average incoming message processing time" on my frontend server is most of the time above the threshold of 1second.

    It is more around 5s with peaks reaching 12s !

    What can cause these high values?

    CPU usage is rarely crossing 50% so I would be surprised if it comes from this.

    Sunday, May 1, 2016 6:58 AM

Answers

  • Hi,

    SfB is real-time media application, when you design the solution you need to consider peak usage not the average. 

    Resource requirements will be varied based on usage pattern of the organization. Microsoft publish HW spec based on the usage profile they have specified.

    If you are deviating from that you need do a stress based on the your organization profile. 

    Since you are already deployed run the performance counters for few weeks and identify where the bottlenecks.

    Increase number of CPU cores, 4 core not sufficient. If you are in shared visualized platform do you have configured minimum reservation for resources,  otherwise other applications will consume resource. 

    In term of disk you may struggling on IOPS. 


    Malinda Perer | MCSM - Chatter Communication | MCM - Lync 2010 and OCS 2007 | http://malinblog.com/ | https://sg.linkedin.com/in/malindap

    • Marked as answer by Eason Huang Sunday, May 15, 2016 11:39 AM
    Thursday, May 12, 2016 1:09 PM
  • When designing SfB/Lync it should be done with the peak usage in consideration.  Just like Malinda stated.  trying to design this with an average usage can cause performance issues at peak times.

    I would recommend in the virtualized environment make sure that your resources are reserved for the server usage. If they are not this can cause an issue when other systems begin to use them and are contending for these resources. 

    I have also seen issues in virtualized environments resulting from back plane issues on the hosts.  If they are having other issues related to the back plane processing, they can present themselves in this and other ways. 

    • Marked as answer by Eason Huang Sunday, May 15, 2016 11:39 AM
    Thursday, May 12, 2016 5:55 PM

All replies

  • Think this could be cause of slow disk performance. Is this setup using virtual machines? If so please check that disk is not set to dynamic, but fixed size.

    Please mark as helpful if you find my contribution useful or as an answer if it does answer your question. That will encourage me - and others - to take time out to help you. Thank you! Off2work

    • Proposed as answer by Eason Huang Thursday, May 12, 2016 9:59 AM
    Sunday, May 1, 2016 8:37 AM
  • Hi,

    Please check if meet the hardware requirement for SFB Server:

    https://technet.microsoft.com/en-us/library/dn951388.aspx

    Also, check if there is any network issue during that time.

    Best Regards


    Please remember to mark the replies as answers if they help, and unmark the answers if they provide no help. If you have feedback for TechNet Support, contact tnmff@microsoft.com.

    Eason Huang
    TechNet Community Support

    • Proposed as answer by Eason Huang Thursday, May 12, 2016 9:59 AM
    Thursday, May 5, 2016 2:40 AM
  • I update ressources on all servers and I am still a bit under the recommendations.

    Frontend  4 CPU / 20GB Ram

    Edge 4 CPU / 8GB Ram

    I have 750users / 1300 clients (because of desk phones) and the recommendations are for 80.000 users !

    The CPU/Ram are far for being used so I do not see a reason to upgrade again.

    The disks are fixed size in vmware.

    The sensors "average incoming message processing time" and "average holding time for incoming messages" is still over the threshold, even if it is a bit better than before.

    I feel the reason of those bad performances is something else.

    Thursday, May 12, 2016 12:07 PM
  • Hi,

    SfB is real-time media application, when you design the solution you need to consider peak usage not the average. 

    Resource requirements will be varied based on usage pattern of the organization. Microsoft publish HW spec based on the usage profile they have specified.

    If you are deviating from that you need do a stress based on the your organization profile. 

    Since you are already deployed run the performance counters for few weeks and identify where the bottlenecks.

    Increase number of CPU cores, 4 core not sufficient. If you are in shared visualized platform do you have configured minimum reservation for resources,  otherwise other applications will consume resource. 

    In term of disk you may struggling on IOPS. 


    Malinda Perer | MCSM - Chatter Communication | MCM - Lync 2010 and OCS 2007 | http://malinblog.com/ | https://sg.linkedin.com/in/malindap

    • Marked as answer by Eason Huang Sunday, May 15, 2016 11:39 AM
    Thursday, May 12, 2016 1:09 PM
  • When designing SfB/Lync it should be done with the peak usage in consideration.  Just like Malinda stated.  trying to design this with an average usage can cause performance issues at peak times.

    I would recommend in the virtualized environment make sure that your resources are reserved for the server usage. If they are not this can cause an issue when other systems begin to use them and are contending for these resources. 

    I have also seen issues in virtualized environments resulting from back plane issues on the hosts.  If they are having other issues related to the back plane processing, they can present themselves in this and other ways. 

    • Marked as answer by Eason Huang Sunday, May 15, 2016 11:39 AM
    Thursday, May 12, 2016 5:55 PM