This discussion has been locked. The information referenced herein may be inaccurate due to age, software updates, or external references.
You can no longer post new replies to this discussion. If you have a similar question you can start a new discussion in this forum.

Response Time and Packet Loss

Hey everyone,

I’m new to using solarwinds and was wondering if I could get some advice. Currently at the company I work for users are complaining about issues using software on a server. I need to be sure that the issue isn’t the network or the server. At the times the users report slowness issues with the software solarwinds report that the CPU and mem are nowhere near taxed. I do however see an issue with Network latency. See graph below.

Capture.PNG

My question is really about how to read this report. Nornal pings from server to client is around 40 to 13ms. This report is specific to the node so I assume that the latency reported here is for packets coming into the server? Is there a way I can see ALL traffic coming into the server to figure out what is causing this latency or where the latency is coming from? Does this signify a bottleneck in the network due to a high volume of data being moved at this time? If so how can I confirm using solarwinds reports?

Thanks for any info you can give!

  • I don't think a direct answer can be given based on what you've provided.  Are the clients on a LAN with the server?  If so, 13-40 ms is pretty high for LAN connectivity.  OTOH, if they are on one or more separate networks from the server, then even 100 ms is not necessarily enough latency to majorly impact the user experience.  Without knowing anything about the software, it's not possible to rule out other factors relating to the application.  Depending on what type of application this is, maybe some of the SAM monitors could help offer some insight into the slowness. 

  • If I read this correctly, it sounds like you want to setup QoE sensors in NPM if you haven't already.  You will setup a server qoe sensor and network qoe sensor.  You will need a specific qoe server to mirror the port the server is connected to, and a monitor for the network ports involved.  This will monitor the server and the network associated with it to see descripencies with the server/network to pinpoint whether or not it is the server or the network that is the bottleneck.  You can read about QoE here: http://www.solarwinds.com/documentation/Orion/docs/SolarWindsPacketAnalysisSensorDeploymentGuide.pdf

    Just from looking at it though, I am pretty sure you are experiencing latency due to tax on the network (whether it be on the access layer or further upstream). Maybe take a look at the switch or upstream switches during that time to see if you notice high cpu utilization, buffer misses, etc.  Sometimes the ASICs in the switches can get congested and sometimes gets punted to the software to get switched/routed.  This would add some latency, but not necessarily packet loss. During that time slice check response times for other servers on that same switch or use the same upstream path and see if you see latency there.

  • Hi tv's frank

    Another option would be to use a trial version of LANGuardian to take a look at the traffic associated with the server, Like with Q0E it uses deep packet inspection but it extracts different meta data which should be useful for troubleshooting. More in the video below.

    Darragh