Parallel Sysplex

Enhancements for z/OS Parallel Sysplex include:

z/OS R11

The AutoIPL support introduced in z/OS V1.10 is extended to multisystem-capable sysplex configurations with active Sysplex Failure Management (SFM) policies in z/OS in a sysplex.

The sysplex failure management (SFM) is designed to use new Base Control Program internal interface (BCPii) services to determine whether an unresponsive system has failed, expedite sysplex recovery by bypassing delay intervals when possible, and automatically reset failed systems without manual intervention. This new function allows SFM to avoid waiting for a period of time before assuming that systems have failed, improve the responsiveness of failure management, avoid operator intervention, and help limit or avoid sysplex-wide slowdowns that can result from single-system failures.

XCF FDI Consistency enforces consistency between the system Failure Detection Interval (FDI) and the excessive spin parameters. This allows system to perform full range of spin recovery actions before it gets removed from the sysplex and avoids false removal of system for a recoverable situation. Also, a new way to specify an operator notification (OPTNOTIFY) relative to the effective FDI is provided, so that you no longer need to calculate the sum of spin loop timeouts to specify the operator notification interval.

With z/OS 1.11, the system default action is ISOLATETIME(0), changed from PROMPT

z/OS R1.12

Removal of "Sick but not dead" systems that have defined themselves as a "Critical Member". This is intended to help reduce the incidence of sysplex-wide problems that can result from unresponsive critical components. GRS is planned to exploit these XCF critical member functions in both ring and star modes. Additionally, GRS will be designed to monitor key tasks and notify XCF if it detects that GRS is impaired.

New CFSTRHANGTIME keyword to allow a structure to be in "hang" condition before corrective action is taken (Stop rebuild, disconnect or terminate connector). This helps avoid sysplex-wide problems that can result from a CF structure that is waiting for timely responses from CF structure connectors.

Improved time it takes for the second and subsequent systems to join a Parallel Sysplex.

New "Display XCF,REALLOCATE" command to report on progress of reallocates

Scalability improvements:

• Support for up to 2047 structures
• Up to 1 TB structures
• new limits of 247 lock structures, 127 serialized list structures, or 255 unserialized list structures

Improved Parallel Sysplex availability with non-disruptive CF dumping capability. z/OS V1.12, in conjunction with z196 servers and Coupling Facility control code (CFCC) Level 17, is designed to capture Coupling Facility (CF) data nondisruptively in some circumstances, allowing the CF to continue operating. The CF uses a pre-staged dump capture area to avoid collateral effects observable by z/OS, such as message time-outs (observed as interface control checks) or loss of connectivity.

z/OS 1.13

XCF provides a simplified set of interfaces for passing messages within a Parallel Sysplex. New services are designed to allow a server to be established to process messages and to send messages across the sysplex without first joining an XCF group. This is intended to make it easier to exploit XCF services for applications that do not require the member management and monitoring provided by the XCF group services interfaces.

The RMF Postprocessor Coupling Facility Activity and the Monitor III CFSYS report can be used to monitor resources associated with the Coupling Facility and CF links. They both have now been extended to indicate channel path details for each of the Coupling over InfiniBand (CIB) link types. This information can help with monitoring and tuning of the Parallel Sysplex.