Migrating to Parallel Sysplex®

Migrating to parallel sysplex
- A Staged Approach

A staged approach is often the most optimal for deploying parallel sysplex. A five-stage approach is a common migration path:

  • Base Sysplex stage sets the foundation for simplified multisystem management by means of exploiting the XCF.
  • Standard Parallel Sysplex stage when you add CFs to the sysplex but no application data sharing is implemented.
  • Parallel Sysplex with data sharing stage when you replicate (clone) your applications across multiple z/OS images to achieve parallelism and application redundancy.
  • Continuous availability stage requires you to advance the third stage to achieve a true 24/7 application availability during both planned and un-planned outages.
  • Disaster recovery stage where you implement disaster recovery solutions to achieve the required level of recovery during the various parallel sysplex deployment stages.
for more info...

IBM Parallel Sysplex® can greatly advantage your IT environment:

  • In a parallel sysplex 1 to 32 z/OS systems are integrated into one multisystem environment or cluster that can employ a specialized Coupling Facility technology to enable many powerful functions including data sharing, distributing work, and balancing system resources within the sysplex.

  • Parallel Sysplex uniquely provides the infrastructure to enable replication of all the components that deliver the application service and allows you to share data across systems in the sysplex. Clients can deploy data sharing based on business and application requirements. Multiple instances of the application can share the data across multiple images in order to avoid outage and achieve 24/7 application availability.

  • It can yield a superior near-linear industry leading scalability to meet growth in high transaction volumes and response times. z/OS images can be dynamically added to a properly configured coupling facility and new CECs can be added to the parallel sysplex. Customers can then grow their IT infrastructure horizontally and incrementally with no disruption. This also provides customers with a great flexibility.

  • Excellent dynamic workload balancing can be achieved in a Parallel Sysplex. Workloads can be automatically routed to the most suitable systems. Many IBM subsystems (DB2, CICS, IMS, MQ, VTAM, TCP/IP, TSO, and JES) exploit this capability to redistribute work across systems, use all spare capacity anywhere in the Parallel Sysplex and achieve the highest resource utilizations.

  • DB2 Data sharing grants you a great flexibility in configuring your system environment:
    • Customer can deploy multiple data sharing groups on the same parallel sysplex. You might assign one group for production data and another group for testing.
    • DB2 subsystem resources can be optimized differently for different sets of applications. Throughout business hours, you can service on-line transactions on one DB2 subsystem, query on a second subsystem, and batch jobs on a third subsystem. Later on and when on-line and query activity diminish, you change buffer pool settings to favor batch on all three DB2 subsystems. The same optimization can be done at z/OS image level.
    • Customers can consolidate separate DB2 subsystems into a single data sharing group to allow applications access to all sets of data. This simplifies managing and accessing the data and reduces IT cost in case of an acquisition or in case one has been constrained to maintain partitioned production DB2 subsystems.
    • One might want to maintain more than one data sharing group in the same server, and even within the same LPAR.

  • Parallel Sysplex can provide a single system image to users, applications, and the network. It also allows system programmers to have a single point of control. This let customers manage their IT environment much more easily.



Functions you need to implement in your operating environment to exploit parallel sysplex® and data sharing:

Time Synchronization - Common Time Source

In order to have the same time (time synchronization) across all members of the sysplex, all the processors in the sysplex need to be attached to a Sysplex Timer or by its replacement Server Time Protocol (STP).

Buffer Data Coherency

Buffer data coherency must be implemented in a Parallel Sysplex so that Cross-Invalidation is handled automatically by the database managers and the Coupling Facility (CF). Each database manager let the CF know every time it adds a record to its local buffer and every time it updates one of those records. Since the CF knows which database managers have a copy of each record, it also knows which instances it has to inform when a given record is updated.

Data access Serialization

Serializing data access is a key requirement in a data sharing parallel sysplex and it is achieved by applying a lock structure in the CF. A database manager instance that needs to access (read or update) a record must first send a lock request to the CF, declaring the record to be accessed and type of the requested access. The CF grants, denies or queues the request depending on its compatibility with other pending updates and requests in progress.

System Monitoring & Communication

The z/OS’s Cross-System Coupling Facility (XCF) is the system component that provides efficient monitoring of the parallel sysplex and checks the availability of its members. Each system updates, every few seconds, the Sysplex Couple Data Set (CDS) with its current time stamp. At the same time, the XCF checks the time stamp of all members of the sysplex to detect for a system failure. It alerts the remaining members if one of them disappears. The Sysplex Failure Management (SFM) function can be used to automatically remove a failed system from the sysplex. The XCF provides also XCF Signaling Services for communication between programs providing work within the same sysplex.

Workload distribution

The ability to present a single system image to the user and to have multiple work managers use the same name combined with z/OS’s Workload Manager (WLM) can provide superior and well balanced workload distribution within the parallel sysplex. WLM works jointly with VTAM, TCP and the transaction managers to determine which system is most equipped for running each portion of work. This is essential in order to attain the key goals of achieving workload’s performance targets, gaining maximum utilization of capacity available anywhere in the sysplex and obtaining 24/7 application availability by shielding users from planned or unplanned outages.

Other z/OS® services to be used in sysplex

Automatic Restart Manager: restarts failed systems and coordinates with WLM to determine which system is most appropriate to restart the failed system on.
CICSPlex: Provides a single point of control for all the CICS regions in the sysplex and also control transaction routing to AOR’s (Application Owning Region).
Global Resource Serialization: to coordinate access to and ensure the integrity of shared resources.
System Logger: to provides a single syslog that merges logs from all the systems.