Active-active replication pioneer Bank-Verlag prefers HP NonStop servers

Longtime HP NonStop server customer Bank-Verlag of Germany almost certainly employed the first instance of “active-active” replication technology—that is, the use of two systems to process transactions concurrently, with each system capable of taking over should the other suffer a technical problem or a natural disaster. As often happens, the inventor, Wolfgang Breidbach, had one need in mind but, as a by-product of his efforts, discovered an even more valuable use for his work.

This intriguing story is well documented by Bill Highleyman, noted enterprise systems consultant and high availability expert, in his new online publication, The Availability Digest. The complete story, excerpted here, is an excellent read…and a subscription to The Availability Digestwell worth your time.

Bank-Verlag was initially the publishing arm of an association of some 300 German banks, including Deutsche Bank and other very large institutions. Wolfgang was hired in 1985 as their first IT employee, to support a new charter to create an online banking service for their smaller banks. His first task was to provide debit cards to be processed on a loaned IBM 370.

Shortly after introduction, a TV exposé showed how easy it was to misuse these new cards based on Bank-Verlag’s first implementation, and soon Wolfgang and colleagues were building a central authorization facility on the same IBM 370.

Due to an acquisition within the group, Bank-Verlag directed Wolfgang to bring up a new independent system for card authorizations. And at this point, ATM use was accelerating. During his research, Wolfgang learned about, and preferred, the more highly available architecture of a newer company, Tandem [Computers] (NonStop [servers]). Moving the data currently on the IBM over to NonStop [servers], without denying user access, was a huge challenge. Thus, Wolfgang and his team began work on what would become the first known use of an active-active system design: They soon had all transactions running on both systems, and rather quickly moved all ATM machines over to the Tandem, leaving the IBM as a hot standby for disaster tolerance. By 1989 ATM use exploded and Bank-Verlag purchased another Tandem (VLX) to handle transactions. Wolfgang used the same strategy to bring up the new system with an even more cost-effective communications design. His team developed their own method for keeping the two databases in sync while sharing the transactional load, much the way that current asynchronous active-active replication detects data collisions needing repair. However, in seventeen years, his system has had not one collision.

This news is published on NonStop Computing websites. January 2007.