Junos OS

> show interfaces e1-0/0/0:26 extensive  
Physical interface: e1-0/0/0:26, Enabled, Physical link is Up
  Interface index: 173, SNMP ifIndex: 79, Generation: 56
  Link-level type: Cisco-HDLC, MTU: 1504, Clocking: Internal, Speed: 1984kbps, Loopback: None, FCS: 16, Framing: G704-NO-CRC4,
  Parent: cau4-0/0/0 Interface index 133
  Device flags : Present Running
  Interface flags: Point-To-Point SNMP-Traps Internal: 0x4000
  Link flags : Keepalives
  Hold-times : Up 0 ms, Down 0 ms
  Keepalive settings: Interval 10 seconds, Up-count 1, Down-count 3
  Keepalive statistics:
  Input : 8070 (last seen 00:00:06 ago)
  Output: 8101 (last sent 00:00:06 ago)
  CoS queues : 4 supported, 4 in use
  Last flapped : 2009-01-20 00:46:06 MSK (6w0d 14:21 ago)
  Statistics last cleared: 2009-03-02 16:42:44 MSK (22:24:59 ago)
  Traffic statistics:
  Input bytes : 1519787574 1593248 bps
  Output bytes : 2282796177 1577520 bps
  Input packets: 5532337 347 pps
  Output packets: 5062043 295 pps
  Input errors:
  Errors: 0, Drops: 0, Framing errors: 0, Runts: 0, Giants: 0, Policed discards: 21514, L3 incompletes: 0, L2 channel errors: 0,
  L2 mismatch timeouts: 0, HS link CRC errors: 0, SRAM errors: 0, Resource errors: 0
  Output errors:
  Carrier transitions: 0, Errors: 0, Drops: 0, Aged packets: 0, MTU errors: 0, Resource errors: 0
  Queue counters: Queued packets Transmitted packets Dropped packets
  0 best-effort 4050648 4033984 25032
  1 gold 1019953 1019909 50
  2 voice 0 0 0
  3 network-cont 8150 8150 0
 
Hi men, I have simple e1 and run extensive output, there I monitore Dropped packects which means "Dropped packets—Number of packets dropped by the ASIC's RED mechanism." I want to avoid it or delete RED. Any advice? 

If you haven't configured a drop-profile, the default is 100% drop probability when the queue is 100% full.  In other words, the output exceeds line-rate for a time period longer than the buffer is able to queue.  RED isn't really doing anything "early".  Or these, might just be tail-drops.

 

Sometimes it isn't obvious, because the line-rate is exceeded for a very short period of time (i.e. microburst in subsecond).  So, when you look at a traffic graphs or monitor the interface, it seems like the output rate is below the max. But, the  period in which the drop happens is too short to "see".

 

The most common solution to alleviate microbursting is to shape the traffic, where it enters the network.  You will commonly see this where the source of the traffic is at FastE (or greater), and the output interface is E1/T1.  Upstream congestion can also trigger microbursting.

 

Regards,

Ben

The problem solved! 

Dropped packets disappeared after increasing buffers.

 

chassis {

  fpc 3 {

  pic 0 {

  q-pic-large-buffer;

 

 http://www.juniper.net/techpubs/software/junos/junos75/swconfig75-cos/html/cos-scheduler-maps7.html#1209142