z/OS MVS Setting Up a Sysplex
Previous topic | Next topic | Contents | Contact z/OS | Library | PDF


Message information for some MVS groups

z/OS MVS Setting Up a Sysplex
SA23-1399-00

This topic provides information on the lengths, frequency, and performance requirements for messages sent by some MVS™ groups. This information can be useful when defining transport classes.
  • Consoles

    For the consoles XCF groups, SYSMCS and SYSMCS2, the length of messages vary. The maximum length of a message is 4K.

  • DAE

    The dump analysis elimination (SYSDAE) group issues one message per dump request, whether or not the dump is suppressed. Each message is approximately 400 bytes long.

  • Global resource serialization
    • RING mode

      For the global resource serialization XCF group, SYSGRS, the most important message to consider is the RSA-message, which holds information about requests for global resources. The RSA-message can contain up to 32K bytes of data. The RSA-message will be sent frequently and will have high performance requirements. (For example, the performance of data base programs and the speed with which access to data sets is granted will depend on the speed with which the RSA-message is processed.) If an alternate serialization method is used, then the performance requirements for global resource serialization signaling are not as critical.

    • STAR mode

      Global resource serialization uses the SYSGRS group for QSCAN (GQSCAN and ISGQUERY), global ENQ contention notification, and Enhanced Contention Analysis (ECA) processing. Signals for these functions might be different in size. The maximum size a signal can be is 61K bytes. If more than 61K bytes of data is required to complete the request, the message is split into more than one part.

    Global resource serialization also creates the SYSGRS2 XCF group in RING and STAR modes but does not use the SYSGRS2 XCF group in XCF communications. Global resource serialization is used in all sysplex configurations even if a different serialization product is used. Global requests that are managed by GRS must be processed immediately.

  • JES2
    The JES2 XCF group issues two kinds of messages:
    • During a sysplex reconfiguration, messages of less than 1K in length will flow regularly between the JES2 XCF group members in the sysplex.
    • As a JES2 system in a sysplex joins a network, it receives JES2 XCF messages with information about the other systems in the network. These messages can be up to 32K bytes in length and contain approximately 30 - 40 bytes of data about each system connected to the network. (Messages longer than 32K, are sent in two or more separate messages.)
  • JES3

    The JES3 group uses XCF messages to transport JES3 staging areas between the JES3 XCF group members in the sysplex. Most staging areas are less than 1K in length; however, the JES3 XCF group combines multiple staging areas and transports them in a single message that can be up to 64K bytes long.

  • WLM

    The WLM group, SYSWLM, sends a message approximately once every ten seconds. The length of the message depends on the number of service classes defined in the WLM policy definition. It is advantageous, but not critical, for the message to be received at once.

    Estimate the size needed as follows: 320 bytes * the number of service class periods (in all service classes) that have a response time or velocity goal.

    The message is sent by MVS images running in GOAL mode to other MVS images running in GOAL mode. If you have not yet migrated two systems to GOAL mode, no messages will be sent.

  • XCF
    The XCF group, SYSXCF, issues Start of change several different End of change kinds of messages:
    Signals requesting data from some other system in the sysplex
    These signals are typically less than 956 bytes.
    Signals responding to such requests
    The length of these responding messages depends on the amount of data to be returned. As used by XCF, the response provides information about the members that reside on the system that is asked to gather the data. If the queried group on such system has only one member, the response is typically less than 956 bytes. For groups that have many members, tens of thousands of bytes of data is possible.
    Signals issued by a system that voluntarily sends member data to other systems in the sysplex
    Such signals are typically less than 956 bytes
    Start of change Signals issued on behalf of an XCF Client/Server service exploiter End of change
    Start of change The sizes of these signals are based on the amount of data the exploiter wishes to send. End of change
    With current usage, the Start of change first three types of End of change signals usually occur infrequently since they are sent in response to the DISPLAY,XCF,GROUP,grpname,memname command. They are also broadcast to the sysplex when an XCF group member is considered stalled. If the stall persists, the signals might reoccur periodically, typically on the order of minutes. If future applications elect to exploit the IXCMG GATHERFROM=OTHER service to collect data from other systems, the frequency, size, and performance requirements can change in response to their usage.

    Start of change The frequency and size of the Client/Server related signals depend on the usage of the service exploiters. End of change

Go to the previous page Go to the next page




Copyright IBM Corporation 1990, 2014