4 Replies Latest reply on May 14, 2010 2:49 PM by GeorgeK

Bandwidth utilization for physical interface vs. subinterface

Steve Pfister May 13, 2010 2:42 PM

In looking at the bandwidth utilization for our upstream boarder router, we're noticing something strange. The connection to our upstream provider is a fiber Gigabit interface which has 2 subinterfaces for our 2 BGP peers (Internet and Internet2). We have all three interfaces added to our NPM 9.51. It looks like the the Internet subinterface is usually about 5 or 6 Mbps higher in utilization than the main interfaces. I would have thought if there was a difference, it would be the other way around. Does anyone know why this is? Is this an Orion thing or a Cisco thing?

Re: Bandwidth utilization for physical interface vs. subinterface

Andy McBride May 13, 2010 4:58 PM (in response to Steve Pfister)

Hmmm. Could be a counter issue. Are all 3 using 64 bit counters?
Like Show 0 Likes(0)

Actions
- Re: Bandwidth utilization for physical interface vs. subinterface
  
  Steve Pfister May 14, 2010 8:31 AM (in response to Andy McBride)
  
  I'm not sure... how would I tell?
  Could it just be a problem with the two interfaces being polled at separate times?
  
  Like Show 0 Likes(0)
  
  Actions
  - Re: Bandwidth utilization for physical interface vs. subinterface
    
    Andy McBride May 14, 2010 9:35 AM (in response to Steve Pfister)
    
    I don't think it is an issue of asynchronous polling as we poll the device with a GETBULK so we get almost everything at the same time. As Admin -> Settings -> Manage Nodes -> Select Nodes in the Show box. Check off the ode of interest and click Edit. Check Allow 64 bit counters box, Submit.
    
    Like Show 0 Likes(0)
    
    Actions
    - Re: Bandwidth utilization for physical interface vs. subinterface
      
      GeorgeK May 14, 2010 2:49 PM (in response to Andy McBride)
      
      I've seen a similar effect with lower-speed interfaces where it's clearly not a 32-bit vs 64-bit issue -- in my case it was DS-1's combined into a multilink, and the sum of the individual interfaces was not exactly the same as the overlying multilink. I also thought it was attributable to asynchronous polling.
      
      Like Show 0 Likes(0)
      
      Actions

Go to original post