IBM Parallel Sysplex
| Industry | Computing, Mainframe technology |
|---|---|
| Founded | 1990 |
| Founder | IBM |
| Products | Parallel Sysplex clustering technology, Coupling Facility hardware, Server Time Protocol, Geographically Dispersed Parallel Sysplex (GDPS) |
| Website | www |
In the realm of computing, a Parallel Sysplex refers to a cluster of IBM mainframe systems functioning as a unified single system image under z/OS. Engineered primarily for disaster recovery, it integrates data sharing with parallel computing techniques, allowing up to 32 systems to distribute workloads efficiently, thereby achieving both high performance and high availability.
Sysplex
Originally introduced by IBM mainframes in 1990 alongside MVS/ESA SPV4.1, a Systems Complex or Sysplex facilitates communication among authorized components within as many as eight logical partitions (LPARs) using the XCF protocol.
Essential elements of a Sysplex include:
- A unified time source that synchronizes system clocks, implemented either through a Sysplex timer (Model 9037) or via the Server Time Protocol (STP)
- Global Resource Serialization (GRS), which manages concurrent resource access by enforcing serialization to ensure exclusive control
- The Cross System Coupling Facility (XCF) that supports peer-to-peer messaging
- Couple Data Sets (CDS)
Prominent Sysplex functionalities encompass:
- Console services that integrate multiple MCS consoles into one cohesive operational interface
- Automatic Restart Manager (ARM), which applies policies to restart failed jobs automatically across LPARs
- Sysplex Failure Manager (SFM), responsible for automatic failure detection and system reconfiguration
- Workload Manager (WLM), which governs diverse workloads across z/OS instances and AIX via policy-based controls
- Enhanced GRS communications utilizing XCF links and Dynamic RNLs
- Tivoli OPC support for hot standby controllers
- RACF commands operable across the entire Sysplex for security administration
- PDSE file sharing capabilities
- Multisystem coordination tools such as VLFNOTE, SDUMP, SLIP, and DAE
- Resource Measurement Facility (RMF) offering reporting across the Sysplex
- CICS improvements in transaction routing and responsiveness by leveraging XCF instead of VTAM
- zFS enabling cross-LPAR data accessibility through XCF
Parallel Sysplex
In April 1994, IBM introduced Parallel Sysplex featuring the 9674Coupling Facility (CF), advanced S/390 hardware models, high-speed coupling links, and MVS/ESA SP V5.1 operating system support.
The CF may operate as a dedicated server running Coupling Facility control code (CFCC), as Internal Coupling Facilities (ICFs) embedded within mainframe processors, or less frequently as standard LPARs. It houses Lock, List, and Cache structures designed to handle serialization, inter-system messaging, and buffer coherence among LPARs.
The core advantage of Parallel Sysplex is its data sharing functionality that permits multiple databases to have simultaneous direct access. This capability enhances:
- The removal of single points of failure at server, LPAR, or subsystem layers
- Application uptime and reliability
- The maintenance of a single system image
- Dynamic load balancing of sessions
- Adaptive transaction routing
- Scalability in processing capacity
Databases supported include:
- IBM Db2
- IBM Information Management System (IMS)
- VSAM (VSARM/RLS)
- IDMS
- Adabas
- DataCom
- Oracle
Resource sharing facilitated by the CF extends to:
- Shared catalogs that boost catalog I/O efficiency
- CICS counter and transient data sharing/recovery
- DFSMShsm workload balancing for efficient data migration
- GRS Star technology that lowers CPU usage and response times during dataset allocation; tape switching enables tape unit sharing across z/OS images
- Dynamic management of CHPID and I/O priorities
- JES2 checkpoint enhancements tailored for multisystem environments
- Consolidation of multisystem logs via Operlog / Logrec
- Shared datasets under RACF simplifying security administration
- WebSphere MQ shared queues enhancing availability and flexibility
- WLM’s Intelligent Resource Director orchestrating CPU and I/O resource allocation across LPARs
Performance optimization with minimized hardware requirements is achieved through multi-system enclave management using XCF Star.
Fundamental components of Parallel Sysplex include:
- Coupling Facility hardware (either CF or ICF) supporting shared data caching and workload distribution
- Synchronization mechanisms such as Sysplex Timers or Server Time Protocol (STP)
- Robust high-speed redundant cabling infrastructure
- Software layers comprising operating systems and middleware like IBM Db2
The CF can be deployed as an external dedicated system or as integral processors within mainframes. For production environments requiring data sharing, deploying at least two CFs or ICFs is advised to ensure redundancy. Since 2005, System z models starting with the z990 have replaced physical Sysplex Timers with STP embedded in mainframe microcode. This advancement allows complete Parallel Sysplex configurations with only two interconnected mainframes or even within a single mainframe setup for testing purposes.
Extensive technical insights on Parallel Sysplex have been documented in IBM Systems Journal publications.
Server Time Protocol
Precise time alignment is essential in transaction-processing environments to guarantee accurate recovery from log data. The Server Time Protocol (STP) delivers a consistent time reference across multiple servers. Drawing on Network Time Protocol concepts, the Hardware Management Console (HMC) designates one System z server as the primary time source (Stratum 1), which then transmits timing signals to Stratum 2 servers via coupling links; these, in turn, forward the signals to Stratum 3 servers. To maintain continuous availability, a secondary time source and an Arbiter server can be assigned to handle failover during anomalies.
STP has been implemented on System z servers since 2005.
Geographically Dispersed Parallel Sysplex
Geographically Dispersed Parallel Sysplex (GDPS) expands the Parallel Sysplex framework across multiple sites, supporting both single-location and multi-location setups:
- GDPS HyperSwap Manager: Utilizes synchronous Peer to Peer Remote Copy (PPRC) within one data center to enable seamless automatic failover between primary and secondary storage without interrupting applications.
- GDPS Metro: Applies synchronous PPRC for distances up to 200 kilometers, allowing two systems to function as a unified entity with automatic recovery and minimal data loss.
- GDPS Global - XRC: Employs asynchronous Extended Remote Copy (XRC) with no distance constraints; recovery involves user initiation but proceeds automatically thereafter, limiting data loss to mere seconds.
- GDPS Global - GM: Based on asynchronous IBM Global Mirror technology designed for complete site failure recovery by activating backup systems and secondary storage.
- GDPS Metro Global - GM and GDPS Metro Global - XRC: Disaster recovery solutions spanning multiple sites that integrate Metro with Global GM or XRC technologies.
- GDPS Continuous Availability: Offers both disaster recovery and continuous availability over unlimited distances by running identical applications and data sets for cross-site workload balancing, supported by IBM Multi-site Workload Lifeline for monitoring and traffic routing.
These solutions enhance enterprise robustness by enabling automated failover, disaster recovery, and workload distribution across geographically separated mainframe infrastructures.