Browse Prior Art Database

Method and Apparatus for Consistent Preservation of Order Dependent Updates for Data Integrity Across Logical Geographic Mirroring

IP.com Disclosure Number: IPCOM000029475D
Original Publication Date: 2004-Jun-30
Included in the Prior Art Database: 2004-Jun-30
Document File: 2 page(s) / 44K

Publishing Venue

IBM

Abstract

The following describes a programming method which allows updates which are not dependent upon one another to be performed concurrently across a remote geographically mirrored set of systems while preserving the sequence of order dependent operations.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 51% of the total text.

Page 1 of 2

Method and Apparatus for Consistent Preservation of Order Dependent Updates for Data Integrity Across Logical Geographic Mirroring

Many applications may make updates that are order dependent. For example, an application may require that an update to data area A must occur before an update to data area B can be reflected. If the updates to the data on disk ever appears out of order, the data may be considered corrupted or damaged. With Geographic Mirroring, asynchronous updates can be made in order to preserve the order of updates, however, unless the updates are quiesced to a known point, the data is unusable on the Mirror Copy since the updates are occurring in parallel to various pieces of data which may be dependant on one another. Any abnormal termination of the Mirror Copy would leave an unusable Mirror Copy. To preserve the order and maintain a useable, coherent Mirror Copy in the event of an abnormal termination, synchronous updates can be made on the Mirror Copy before control is returned to the Production Copy. There are several ways to achieve this. As an operation is issued on the Production Copy, the same operation can be issued to the Mirror Copy and all other operations could wait for the operation to complete on both copies before being issued. This has the advantage of preserving the order of updates on both systems, however, it is extremely slow in that all operations are now single threaded to both systems. Another alternative is to bundle a set of operations. At any point in time, there are any number of outstanding operations which are occurring in parallel (i.e., multiple writes to multiple disk arms). A snapshot can be made of these operations which are occurring in parallel and the bundle of the operations in the snapshot could be sent to the Mirror Copy to be performed. Within the bundle, operations could be performed in any order on the Mirror Copy since they were occurring in parallel on the Production Copy, however, operations within the bundle on the Mirror Copy would need for all of the operations within a bundle to complete before proceeding to the next bundle. This also has the advantage of preserving the order of dependent operations since in general there is no dependancy on the order of the operations which occur within the bundle since they can occur in any order on either system. All operations which are dependent on one another would occur in different bundles by d...