General TCP/IP Problems

There are many causes of general TCP/IP problems.

This table lists TCP/IP problems and factors to consider:

Problem Factors to Consider
Host Issues
  • Inadequate memory
  • Slow disk speed/contention
  • Slow channel speed/contention
  • Excessive workload
  • Inadequate processors/slow processor speed
  • Inefficient performance groups and dispatch priorities
  • Resource competition among applications on same system
  • Resource competition among LPARs
Network issues At the link level, look for:
  • Link errors
  • Hardware or interface errors
  • Latency problems
  • Collisions

At the IP layer, look for:

  • Discarded packets
  • Reassembly failures
  • Whether the DoNotFragment Bit is set
  • TOS (TCP/IP Type of Service) such as Telnet with low delay (interactive priorities overriding batch transmissions)
  • Small MTU/MSS (Maximum Transmission Unit/Maximum Segment Size)
    Note: If the MTU is too small, inefficiency results whereas if it is too large, datagram fragmentation may result.
Network issues (cont'd) At the TCP layer, look for:
  • Segments retransmitted
  • Connections reset
  • Frequency of ACKs
  • Window size too small

In the TCP/IP stack, look for:

  • The maintenance level of the two TCP stacks involved
  • The use of PORT and 1364 TCP CDSTC NODELAYACKS which may delay ACKS
  • The values of TCPSENDBFRSIZE and TCPRCVBFRSIZE in the TCP/IP PROFILE data set
    Note: These values affect all applications using the TCP/IP protocol whereas the V2.BUFSIZE initialization parameter (see below) affects the operation of the IBM® Connect:Direct® application only.
  • Whether the value set for PATHMTUDISCOVERY is an MTU size of 8992 (the default)

In the IBM Connect:Direct global initialization parameters file, look at: