IBM Parallel Sysplex


In computing, a Parallel Sysplex is a cluster of IBM mainframes acting together as a single system image with z/OS. Used for disaster recovery, Parallel Sysplex combines data sharing and parallel computing to allow a cluster of up to 32 systems to share a workload for high performance and high availability.

Sysplex

In 1990, IBM mainframe computers introduced the concept of a Systems Complex, commonly called a Sysplex, with MVS/ESA SPV4.1. This allows authorized components in up to eight logical partitions to communicate and cooperate with each other using the XCF protocol.
Components of a Sysplex include:
Users of a Sysplex include:
The Parallel Sysplex was introduced with the addition of the Coupling Facility with coupling links for high speed communication, with MVS/ESA V5.1 operating system support, together with the mainframe models in April 1994.
The Coupling Facility may reside on a dedicated stand-alone server configured with processors that can run Coupling Facility control code, as integral processors on the mainframes themselves configured as ICFs, or less common, as normal LPARs. The CF contains Lock, List, and Cache structures to help with serialization, message passing, and buffer consistency between multiple LPARs.
The primary goal of a Parallel Sysplex is to provide data sharing capabilities, allowing multiple databases for direct reads and writes to shared data. This can provide benefits of
Databases running on the System z server that can take advantage of this include:
Other components can use the Coupling Facility to help with system management, performance, or reduced hardware requirements. Called “Resource Sharing”, uses include:
Tape Switching uses the GRS structure to provide sharing of tape units between z/OS images.
Multi-system enclave management for improved performance
Major components of a Parallel Sysplex include:
The Coupling Facility may be either a dedicated external system or integral processors on the mainframes themselves configured as ICFs. It is recommended that at least one external CF be used in a parallel sysplex. It is recommended that a Parallel Sysplex has at least two CFs and/or ICFs for redundancy, especially in a production data sharing environment. Server Time Protocol replaced the Sysplex Timers beginning in 2005 for System z mainframe models z990 and newer. A Sysplex Timer is a physically separate piece of hardware from the mainframe, whereas STP is an integral facility within the mainframe's microcode.
With STP and ICFs it is possible to construct a complete Parallel Sysplex installation with two connected mainframes. Moreover, a single mainframe can contain the internal equivalent of a complete physical Parallel Sysplex, useful for application testing and development purposes.
The IBM Systems Journal dedicated a full issue to all the technology components.

Server Time Protocol

Maintaining accurate time is important in computer systems. For example, in a transaction-processing system the recovery process reconstructs the transaction data from log files. If time stamps are used for transaction-data logging, and the time stamps of two related transactions are transposed from the actual sequence, then the reconstruction of the transaction database may not match the state before the recovery process.
Server Time Protocol can be used to provide a single time source between multiple servers. Based on Network Time Protocol concepts, one of the System z servers is designated by the HMC as the primary time source. It then sends timing signals to the Stratum 2 servers through use of coupling links. The Stratum 2 servers in turn send timing signals to the Stratum 3 servers. To provide availability, one of the servers can be designated as a backup time source, and a third server can be designated as an Arbiter to assist the Backup Time Server in determining if it should take the role of the Primary during exception conditions.
STP has been available on System z servers since 2005.
More information on STP is available in “Server Time Protocol Planning Guide”.

Geographically Dispersed Parallel Sysplex

Geographically Dispersed Parallel Sysplex is an extension of Parallel Sysplex of mainframes located, potentially, in different cities. GDPS includes configurations for single site or multiple site configurations: