International Business Machines Corporation

Pierre-Fortin Jean-Denis

Gail Andrea Spear

Robert Francis Bartfai

Sonny Earl Williams

Warren Keith Stanley

Nicolas Marc Clayton

Sherry Suzanne Worcester

Provided are a method, system, deployment and program for managing data transfer in consistency group formation. In one embodiment, nodes are periodically polled to determine traffic conditions. An estimated drain time for a node such as the most congested node, for example, is compared to a drain time limit and if within a vicinity of the drain time limit, operations switch to a consistency group formation mode. Other embodiments are described and claimed.

Node polling in consistency group formation

Provided are a method, system, deployment and program for managing data transfer in consistency group formation. In one embodiment, a transfer time for a data structure is selected as a function of at least one determined backlog. Upon expiration of the selected transfer time, the mode of operation may be switched. By dynamically selecting a transfer time, backlogs may be reduced to increase the likelihood of successful creation of a consistency group. Other embodiments are described and claimed.

Data transfer management in consistency group formation

Resynchronization of data between a primary (production) data site and a secondary (recovery) site following a failure is enhanced when the size of a data track at the production site is different from the size of a data track at the recovery site. The recovery site reads an out-of-sync (OOS) bitmap created at the production site and expands or contracts the bitmap to accommodate the size difference. The resulting production site bitmap is merged with a OOS bitmap maintained at the recovery site to indicate those tracks which are to be transferred from the recovery site to the production site. Thus, only those tracks which are required to be transferred are transferred. Buffer space may be allocated in which to expand or contract the production site OOS bitmap. Buffer space may be conserved by sequentially reading portions of the production site OOS bitmap into a small buffer. In a first operation, half of the bitmap portion is discarded and the remaining half is expanded into the buffer and merged with a corresponding section of the recovery site OOS bitmap. In a second operation, the portion is re-read into the buffer and the other half discarded. The remaining half is then expanded into the buffer and merged with the corresponding section of the recovery site OOS bitmap. The process continues until the entire production site OOS bitmap has been merged with the recovery site OOS bitmap.

Growing community of inventors

Tucson, AZ, United States of America

Pierre-Fortin Jean-Denis