IBM i, iSeries, IBM Power Systems Technical Consultant
(This new solution is purely for DR – for high availability, you still need to look to IBM PowerHA clustering technology.)
With the introduction of GDR for IBM i, IBM is expanding on the current DR solutions available for IBM i. In particular, the solution leverages external storage and virtualization to help insure that the i platform remains a viable, up-to-date element in the IBM ecosystem.
In a nutshell, GDR enables IBM i LPARs to be remotely restarted on other IBM Power Systems servers in a very easy and affordable way. The LPARs involved in the GDR remote restart must use SAN storage, including IBM DS8000, SAN Volume Controller (a.k.a. Spectrum Virtualize), Storwize or EMC storage arrays. To provide even more robust DR capability, the storage arrays can be geographically dispersed.
GDR leverages existing storage replication technologies such as IBM Global Mirror (asynchronous), Metro Mirror (synchronous) and EMC’s SRDF. The solution uses a number of concepts and technologies from Live Partition Mobility (LPM). This will be familiar to most larger IBM i shops, which are likely already to use IBM Virtual I/O Server to virtualize storage and network resources. The main difference between GDR and LPM is that LPM does not reboot LPARs when moving them to a remote Power server – rather, it pauses and moves the resources. In contrast, using GDR always includes an abnormal IPL of the LPAR on the remote server.
Since GDR leverages storage replication (using Metro Mirror, Global Mirror or EMC SRDF), it can deliver effectively zero RPO if used with synchronous data replication and near-zero RPO with asynchronous replication. Because an abnormal IPL of the LPAR is required, RTO will vary depending on the size and structure of the database.
GDR for i offers both practical and financial advantages compared with conventional approaches to backup and recovery. It’s better integrated and it works within your supported i landscape rather than being dependent on third-party software. And because it sits on top of established, automated storage replication, minimal management overhead is required to verify that the replication is up to date. Particularly for smaller shops that don’t have huge numbers of skilled i technicians to dedicate to daily administration, cutting administration could be a real boon. Finally, GDR comes as a licence for a single partition (the AIX LPAR called K-SYS that controls the solution). Compared with a conventional Disaster Recovery solution that is likely to be priced by the core, significant cost savings may be achievable.
GDR’s use of an AIX LPAR to manage the switching for IBM i LPARs may be off-putting for some i shops. It will be interesting to see how comfortable they are in adopting the solution, since Virtual I/O Server virtualization (which runs on an AIX LPAR) is probably still not always well understood, especially in small and midrange environments.
If you like the sound of GDR, but have some reservations about the AIX side of things, Northdoor has the required skills in both technologies and would be happy to help.