IBM CIP Protocols

Document ID: 17534

Contents

Introduction
Before You Begin
      Conventions
      Prerequisites
      Components Used
Problem Identification
      Determine If Router Is Fast Switching or Process Switching
      Identify Problem When CIP Is Process Switching Packets
      Identify Problems When CIP Is Fast Switching Packets
LLC2 Tuning Parameters
Conclusion
Related Information

Introduction

With the introduction of Cisco Systems Network Architecture (CSNA) on the Channel Interface Processor (CIP), Cisco 7000/7500 routers are required to provide Systems Network Architecture (SNA) support for more connections, sessions, and data traffic than ever before.

Therefore, the router may not be capable of providing the desired SNA support with the default tuning parameters. Because the CSNA feature can switch packets from the CIP to the Routing Processor (RP) at a high speed, it's possible that packets will be dropped using the default tuning parameters. The router can adjust and recover from packet drops if the frequency is low; however, session loss may occur when the frequency of drops is high.

The purpose of this document is to:

• Help CSNA users identify problem areas with CSNA performance or session loss, or both.

• Provide guidelines for tuning router parameters and identifying problems.

Before You Begin

Conventions

For more information on document conventions, see the Cisco Technical Tips Conventions.

Prerequisites

There are no specific prerequisites for this document.

Components Used

This document is not restricted to specific software and hardware versions.

Problem Identification

If you are experiencing problems with inadequate performance or session loss, or both, in the CSNA environment, you can examine trouble areas within the router to determine if tuning can help.

Determine If Router Is Fast Switching or Process Switching

Problems within the CSNA router can be isolated depending on the type of packet switching with which the CIP is primarily involved. The CIP supports two types of switching:

• Fast switching.

• Process switching.

Fast switching allows the CIP to switch packets directly to another interface processor without queuing the packet to the RP. Process switching is slower because the CIP must copy the packet into a buffer and queue the buffer for processing by the RP. It is necessary to isolate problems differently depending on whether the CIP is fast switching or process switching. Often, it can be determined whether the CIP is fast switching or process switching packets based on the router features configured and used.

Several router features are always process switched:

• Advanced Peer-to-Peer Networking (APPN)

• Data-Link Switching Plus (DLSw+)/TCP

• Downstream Physical Unit (DSPU)/RSRB

• Remote Source-Route Bridging (RSRB)/TCP

• RSRB/TCP with local-ack

• Synchronous Data Link Control Logical Link Control (SDLLC)

Other router features are always fast switched:

• DLSw+/direct

• DLSw+/Fast-Sequenced Transport (FST)

• RSRB/direct

• RSRB/FST

• Source-Route Bridging (SRB)

• Source-Route Translational Bridging (SR/TLB) (as of IOS 11.2)

The definitive method for determining the CIP packet switching type is to issue the following command:

show interface Channel x/2 stats

Example.

dspu-7k#show interface channel 4/2 stats

Channel4/2
          Switching path    Pkts In   Chars In   Pkts Out  Chars Out
               Processor      35126    1270810      35110    1108845
             Route cache          0          0          0          0
          Autonomous/SSE          0          0          0          0
                   Total      35126    1270810      35110    1108845

In the show interface channel command output:

• The Processor counts indicate the number of process-switched packets on the interface.

• The Route cache counts indicate the number of fast-switched packets on the interface.

• The Autonomous/SSE counts indicate the number of autonomously-switched packets on the interface.

  Note: CSNA does not support autonomous packet switching.

Identify Problem When CIP Is Process Switching Packets

Problems with performance or session loss, or both, are most apparent when process switching large numbers of sessions or high data transfer rates, or both.

There are several problem areas in the process switch path that are most common:

1. Buffer misses/failures

2. Input queue overflow on an interface

3. Output queue overflow on an interface

Each of these areas is discussed in a separate document.

Identify Problems When CIP Is Fast Switching Packets

Problems with performance or session loss, or both, when the CIP is fast switching should be rare, but may occur under certain conditions. When packets are dropped during fast switching, the show interface command output shows increasing numbers of output queue drops that cannot be accounted for by process switching drops.

You can tune the TX queue limit for the interface using the tx-queue-limit interface configuration command. The show controller cxbus command output shows the current TX Queue Limit for each interface.

Example.

dspu-7k#show controller cxbus
 
Switch Processor 5, hardware version 11.1, microcode version 10.12
  Microcode loaded from system 
  512 Kbytes of main memory, 128 Kbytes cache memory
  16 256 byte buffers, 4 1024 byte buffers, 51 1520 byte buffers, 38 2068 byte buffers, 73 4496 byte buffers
  Restarts: 0 line down, 0 hung output, 0 controller error
 CIP 4, hardware version 4.1, microcode version 180.40
  Microcode loaded from flash nightly/cip180-40.OffCTA_950922
  Controller Sync: 5 timeouts, 5 resyncs 0 failures, 0 max phase count
  EPROM version 165.0, VPLD version 4.28
  CPU utilization  39%, sram 848/512K, dram 59542704/64M
  Interface 32 - Channel 4/0
    10 buffer RX queue threshold, 18 buffer TX queue limit, buffer size 4496
    ift 0007, rql 5, tq 0000 0000, tql 18
    Transmitter delay is 0 microseconds
  Interface 33 - Channel 4/1
    10 buffer RX queue threshold, 18 buffer TX queue limit, buffer size 4496
    ift 0007, rql 6, tq 0000 0000, tql 18
    Transmitter delay is 0 microseconds
  Interface 34 - Channel 4/2
    10 buffer RX queue threshold, 18 buffer TX queue limit, buffer size 4496
    ift 0007, rql 5, tq 0000 06D8, tql 18
    Transmitter delay on all interfaces is 0 flags

Note: All packets (explorers) must be process switched at least once before fast switching takes over.

Note: If the nature of CSNA problems is with establishing connections, consider isolating problems with process switching rather than fast switching

LLC2 Tuning Parameters

The following Logical Link Control, type 2 (LLC2) session characteristics can be tuned using the llc2 configuration command:

Note: Each adapter under the CSNA internal LAN is represented by a separate LLC2 that may be tuned independently from the other adapters.

Example.

!
interface Channel4/2
 no ip address
 no keepalive
 max-llc2-sessions 1000
 lan TokenRing 0
  source-bridge 666 1 99
  adapter 0 4000.b0ca.7000
   llc2 t1-time 3000
  adapter 1 4000.b0ca.7001
   llc2 t1-time 5000
  adapter 2 4000.b0ca.7002
   llc2 t1-time 10000

Note: When the RSRB/TCP local-ack feature is configured, the additional LLC2 representing LAK may also require tuning at the channel virtual interface level.

Example.

!
interface Channel4/2
 no ip address
 no keepalive
 llc2 t1-time 5000           

!--- Tuning for RSRB/TCP local-ack LLC2.

 max-llc2-sessions 1000
 lan TokenRing 0
  source-bridge 666 1 99
  adapter 0 4000.b0ca.7000
   llc2 t1-time 3000         

!--- Tuning for CSNA adapter LLC2.

Conclusion

The first step of router tuning is to identify the problem area and determine what to tune. Using snapshots of the router's show command output helps to highlight the problem areas.

A single snapshot may identify a particular problem area from the past that overshadows the current problem area.

Example.

In the following show buffer output, it may seem that the primary problem area is Big buffers:

dspu-7k#show buffer

Buffer elements:
     500 in free list (500 max allowed)
     234 hits, 0 misses, 0 created
 
Public buffer pools:
Small buffers, 104 bytes (total 25, permanent 10):
     5 in free list (5 min, 10 max allowed)
     1359 hits, 28 misses, 17 trims, 32 created
     28 failures (0 no memory)
Middle buffers, 600 bytes (total 90, permanent 90):
     90 in free list (10 min, 200 max allowed)
     171 hits, 0 misses, 0 trims, 0 created
     0 failures (0 no memory)
Big buffers, 1524 bytes (total 341, permanent 90):
     120 in free list (5 min, 300 max allowed)
     113128 hits, 1323 misses, 959 trims, 1210 created
     1318 failures (0 no memory)
VeryBig buffers, 4520 bytes (total 10, permanent 10):
     10 in free list (0 min, 300 max allowed)
     20 hits, 0 misses, 0 trims, 0 created
     0 failures (0 no memory)
Large buffers, 5024 bytes (total 10, permanent 10):
     10 in free list (0 min, 30 max allowed)
     0 hits, 0 misses, 0 trims, 0 created
     0 failures (0 no memory)
Huge buffers, 18024 bytes (total 2, permanent 0):
     0 in free list (0 min, 13 max allowed)
     2 hits, 2 misses, 0 trims, 2 created
     0 failures (0 no memory)

The following output shows that the Big buffers are stable and the Small buffer pool is the current problem area:

dspu-7k#show buffer 

Buffer elements:
     500 in free list (500 max allowed)
     234 hits, 0 misses, 0 created
 
Public buffer pools:
Small buffers, 104 bytes (total 42, permanent 10):
     7 in free list (5 min, 10 max allowed)
     1576 hits, 35 misses, 17 trims, 39 created
     34 failures (0 no memory)
Middle buffers, 600 bytes (total 90, permanent 90):
     90 in free list (10 min, 200 max allowed)
     171 hits, 0 misses, 0 trims, 0 created
     0 failures (0 no memory)
Big buffers, 1524 bytes (total 341, permanent 90):
     120 in free list (5 min, 300 max allowed)
     113128 hits, 1323 misses, 959 trims, 1210 created
     1318 failures (0 no memory)
VeryBig buffers, 4520 bytes (total 10, permanent 10):
     10 in free list (0 min, 300 max allowed)
     20 hits, 0 misses, 0 trims, 0 created
     0 failures (0 no memory)
Large buffers, 5024 bytes (total 10, permanent 10):
     10 in free list (0 min, 30 max allowed)
     0 hits, 0 misses, 0 trims, 0 created
     0 failures (0 no memory)
Huge buffers, 18024 bytes (total 2, permanent 0):
     0 in free list (0 min, 13 max allowed)
     2 hits, 2 misses, 0 trims, 2 created
     0 failures (0 no memory)

Related Information

• Technical Support - Cisco Systems