Solved: how to reduce the output drops on interface

roger.malik · ‎05-13-2008

we have connection between 6509 switch interface and 3725 access device. however on output drops are increasing 6509 interface, any ideas how can i make to zero. below are sh int output for both device.

6509-CORE-1-1#sh int gi3/2

GigabitEthernet3/2 is up, line protocol is up (connected)

Hardware is C6k 1000Mb 802.3, address is 0015.2bb1.87d9 (bia 0015.2bb1.87d9)

Description: ::ge3/2::oper::3725-ACCESS-1-1::fa0/0::100M

MTU 1500 bytes, BW 100000 Kbit, DLY 10 usec,

reliability 255/255, txload 17/255, rxload 1/255

Encapsulation ARPA, loopback not set

Keepalive set (10 sec)

Full-duplex, 100Mb/s

input flow-control is off, output flow-control is on

Clock mode is auto

ARP type: ARPA, ARP Timeout 04:00:00

Last input 00:00:33, output 00:00:02, output hang never

Last clearing of "show interface" counters 00:36:44

Input queue: 0/2000/0/0 (size/max/drops/flushes); Total output drops: 4417

Queueing strategy: fifo

Output queue: 0/40 (size/max)

5 minute input rate 6000 bits/sec, 5 packets/sec

5 minute output rate 7013000 bits/sec, 1561 packets/sec

12989 packets input, 2170281 bytes, 0 no buffer

Received 993 broadcasts (993 multicasts)

0 runts, 0 giants, 0 throttles

0 input errors, 0 CRC, 0 frame, 0 overrun, 0 ignored

0 watchdog, 0 multicast, 0 pause input

0 input packets with dribble condition detected

3523848 packets output, 1963109904 bytes, 0 underruns

0 output errors, 0 collisions, 0 interface resets

0 babbles, 0 late collision, 0 deferred

0 lost carrier, 0 no carrier, 0 PAUSE output

0 output buffer failures, 0 output buffers swapped out

3725-ACCESS-1-1#sh int fa0/0

FastEthernet0/0 is up, line protocol is up

Hardware is Gt96k FE, address is 0011.9267.d7c0 (bia 0011.9267.d7c0)

Description: ::fa0/0::oper::6509-CORE-1-1::ge3/2::100M

Internet address is 10.200.0.5/24

MTU 1500 bytes, BW 100000 Kbit, DLY 100 usec,

reliability 255/255, txload 1/255, rxload 1/255

Encapsulation 802.1Q Virtual LAN, Vlan ID 1., loopback not set

Keepalive set (10 sec)

Full-duplex, 100Mb/s, 100BaseTX/FX

ARP type: ARPA, ARP Timeout 04:00:00

Last input 00:00:00, output 00:00:00, output hang never

Last clearing of "show interface" counters 4w5d

Input queue: 0/75/0/0 (size/max/drops/flushes); Total output drops: 0

Queueing strategy: fifo

Output queue: 0/40 (size/max)

5 minute input rate 10000 bits/sec, 9 packets/sec

5 minute output rate 9000 bits/sec, 7 packets/sec

160141872 packets input, 3554767197 bytes

Received 4856018 broadcasts, 0 runts, 0 giants, 0 throttles

0 input errors, 0 CRC, 0 frame, 0 overrun, 0 ignored

0 watchdog

0 input packets with dribble condition detected

154246689 packets output, 3477017281 bytes, 0 underruns

0 output errors, 0 collisions, 0 interface resets

0 babbles, 0 late collision, 0 deferred

0 lost carrier, 0 no carrier

0 output buffer failures, 0 output buffers swapped out

Joseph W. Doherty · ‎05-14-2008

Complete elimination of drops can be impossible to obtain unless the path is not oversubscribed. Most TCP protocol implementations use drops as a primary indication the path has been saturated. Likely the best you will obtain is shifting the (most?) drops to the 3725.

Not sure what the 6500 will support, but you might be able to go much, much larger on the FIFO queue size. Recall, at least on some routers, queue size up to 4K packets supported. (This, even if it does eliminate the drops on the 6500, might just introduce needless latency.)

Also, if supported, you might try RED/WRED. Its max queue size uses a running average, i.e. should pass bursts yet should also drop packets if there's sustained congestion.

View solution in original post

muca · ‎05-14-2008

Output drops are caused by a congested interface.

The number of packets dropped because the output queue is full

Check this link

http://www.cisco.com/en/US/products/hw/switches/ps700/products_tech_note09186a008015bfd6.shtml#shwop

roger.malik · ‎05-14-2008

Hi Muca,

wat steps should i take to reduce the congestion or make output queue free.

thanks agian for your response

Joseph W. Doherty · ‎05-14-2008

Input queue: 0/2000/0/0 (size/max/drops/flushes); Total output drops: 4417

3523848 packets output, 1963109904 bytes, 0 underruns

Your drop percentage is only about .1%; often acceptable. However, you might try, if supported, increasing the outbound FIFO queue size slightly, perhaps about 60 instead of the default 40.

roger.malik · ‎05-14-2008

thanks Joseph,

i did increase the size to till 100 increamently, but it produce not good enough result.

15 is often acceptable, i need to this to completely vanished, is their any other way to resolved it. thanks again.

Joseph W. Doherty · ‎05-14-2008

Complete elimination of drops can be impossible to obtain unless the path is not oversubscribed. Most TCP protocol implementations use drops as a primary indication the path has been saturated. Likely the best you will obtain is shifting the (most?) drops to the 3725.

Not sure what the 6500 will support, but you might be able to go much, much larger on the FIFO queue size. Recall, at least on some routers, queue size up to 4K packets supported. (This, even if it does eliminate the drops on the 6500, might just introduce needless latency.)

Also, if supported, you might try RED/WRED. Its max queue size uses a running average, i.e. should pass bursts yet should also drop packets if there's sustained congestion.