It used to be that GDPS (Geographically Dispersed Parallel Sysplex) was only for the largest mainframe datacenters, and of them mainly for banks. In fact the finance sector accounts for 450 of the 620 GDPS installs to date. The latest enhancements to GDPS, however, open GDPS to a broader set of the systems through support for the zBX and the hybrid, open systems environment.
And GDPS is no longer just for the largest enterprises only. IBM has started to see smaller organizations looking to protect their systems and data with GDPS. To date five SMBs are noted on IBM’s latest GDPS tally. This makes sense with the z114 starting at $75k. Check out this graphic of the GDPS family of products here. Although GDPS does not yet offer something for everyone there are a lot more options than there used to be.
Here is the current GDPS menu of options:
- GDPS/PRRC HM for continuous availability of data within a single data center with an RPO and RTO of zero
- GDPS/PPRC for continuous availability and disaster recoverability within a metropolitan region with an RPO of zero and an RTO of minutes (active/active configuration)
- GDPS/GM& GDPS/XRC for disaster recovery at an extended distance with an RPO of seconds and RTO of less than 1 hr.
- GDPS/MGM & GDPS/MzGM for continuous availability regionally and disaster recovery at extended distance with, depending on the configuration an RPO of zero or seconds and an RTO in minutes
- GDPS/Active-Active for continuous availability, disaster recovery, and cross-site workload balancing at an extended distance with the RPO and RTO in seconds
To get this variety GDPS mixes a variety of replication and mirroring technologies, both synchronous and asynchronous, using such capabilities as HyperSwap and Metro Mirror. HyperSwap, managed through GDPS, is designed to swap a large number of devices and, as IBM puts it, do it fast and with minimal impact to application availability and where disruptions are measured in seconds rather than hours. Surprisingly, HyperSwap can function even if the primary disk subsystem is not operational, which means the organization can survive a primary disk subsystem failure or a complete site failure without recycling the systems. In effect, the organization maintains continuous access to data should such a failure occur.
IBM considers GDPS’s active/active continuous availability (GDPS/AA) as a fundamental paradigm shift from a failover model to a near continuous availability model. This results from unlimited distance replication with only seconds of user impact if there is a site disaster. By using software based replication and advanced techniques for copying the data between sites, GDPS allows control over which workloads are being protected.
Once considered a degree of difficulty bordering on brain surgery, IBM has simplified GDPS with automation that helps manage the availability of the workload and acts as the controller. Also, previously considered a mainframe-only game, GDPS now can handle a hybrid (heterogeneous) systems environment through z/OS using the IBM TotalStorage Enterprise Storage Server. That means shops deploying the zBX and its various blades and workloads can participate in GDPS protection and recovery. GDPS’s hybrid capabilities for the zBX have been available since October with more capabilities introduced this month and more coming this summer. For environments that share disk subsystems between the z/OS and open system platforms, GDPS/PPRC can manage the PPRC for open systems storage and provide data consistency across z/OS and open systems data, according to IBM.
IBM also notes that GDPS/PPRC and GDPS/XRC provide a coordinated DR offering for organizations running a multi-tiered architecture with storage subsystems shared between z/OS and Linux on z. This might be an SAP application server running on Linux on z and the SAP DB server running on z/OS.
GDPS is a powerful data and system protection and recovery product set. For most companies, however, it probably remains costly overkill. That may begin to change with the z114 and extending GDPS to hybrid z-open system environments.