Output queue drops when no overutilization is present.

[momchilandonov]

Hi to all!

I have a Cisco router C887 which is having many discards on the output queue. It is having 1 megabit upload speed ADSL line(ATM interface).

The discards occur at 15-40% utilization of the line.

#sh int atm 0
ATM0 is up, line protocol is up
  Hardware is MPC ATMSAR, address is
  Description: [cewan-phy]
  MTU 1600 bytes, sub MTU 1600, BW 888 Kbit/sec, DLY 410 usec,
     reliability 255/255, txload 1/255, rxload 1/255
  Encapsulation ATM, loopback not set
  Keepalive not supported
  Encapsulation(s): AAL5
  4 maximum active VCs, 1024 VCs per VP, 1 current VCCs
  VC Auto Creation Disabled.
  VC idle disconnect time: 300 seconds
  Last input never, output 00:00:00, output hang never
  Last clearing of "show interface" counters 2w3d
  Input queue: 0/75/0/0 (size/max/drops/flushes); Total output drops: 11085
  Queueing strategy: Per VC Queueing
  5 minute input rate 2000 bits/sec, 3 packets/sec
  5 minute output rate 2000 bits/sec, 3 packets/sec
     2781747 packets input, 907169347 bytes, 0 no buffer
     Received 0 broadcasts (0 IP multicasts)
     0 runts, 0 giants, 0 throttles
     0 input errors, 0 CRC, 0 frame, 0 overrun, 0 ignored, 0 abort
     3148625 packets output, 567004219 bytes, 0 underruns
     66 output errors, 0 collisions, 5 interface resets
     0 unknown protocol drops
     0 output buffer failures, 0 output buffers swapped out

#sh int sum

 *: interface is up
 IHQ: pkts in input hold queue     IQD: pkts dropped from input queue
 OHQ: pkts in output hold queue    OQD: pkts dropped from output queue
 RXBS: rx rate (bits/sec)          RXPS: rx rate (pkts/sec)
 TXBS: tx rate (bits/sec)          TXPS: tx rate (pkts/sec)
 TRTL: throttle count

  Interface                   IHQ       IQD       OHQ       OQD      RXBS      RXPS      TXBS      TXPS      TRTL
-----------------------------------------------------------------------------------------------------------------
* ATM0                          0         0         0               11085      1000         2      1000         2         0

What commands can I do to the router to prevent those discards from happening? Notice that this issue is happening over the last several weeks/months and router reboots and clearing of counters doesn't help. It seems that discards don't happen if the utilization is less than 10% (100 kbps).

I've read that the TCP window change could solve this issue but I am not sure what value do I need to set and what is the command. Do I need to enter the ATM interface and then configure it?

This is not an ADSL specific issue IMO because the SNR was already increased and the line is tested to be okay.

Is it likely an increase of the upload line speed to 2 mbps would solve this issue given that it isn't overutilization?

[Joseph W. Doherty]

Disclaimer

The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.

Liability Disclaimer

In no event shall Author be liable for any damages wha2tsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.

Posting

Yes, setting TCP window size, specifically the receiver's RWIN, might avoid TCP drops.  The ideal setting would be to set RWIN for the BDP between the sender and the receiver, but if you have multiple flows using the link concurrently and dynamically, it would be practically impossible for you to set RWIN.  (NB: some traffic shaping appliances might be able to work this "magic".

What else might you do?  Well, you might start by insuring your egress queuing can handle about half the RWIN.  You can also shape for your upstream bandwidth (if less than port bandwidth) and you manage congestion, if any.  You're probably still going to need to drop packets, but by managing congestion you can prefer dropping some traffic's packets over another's and depending how you drop, you might minimize the overall drop rate.  (NB: realize some protocols, such as TCP, use drops to adjust their transmission rate.)

Is it likely an increase of the upload line speed to 2 mbps would solve this issue given that it isn't over utilization?

Depending on what you traffic is doing, increasing your bandwidth might help or you might even see your drops increase.  However, more bandwidth can usually help if you're actively managing congestion,

BTW, 15 to 40%, alone, doesn't always mean there's not any over utilization.  You might want to research the term "micro burst".