Our little group within NCR is doing practical research (directly leading to product development) in the area of clustering and high availability technologies as they apply to NT Server Systems. While we are looking at all of the "normal" clustering questions, the focus is on minimizing the impact of proprietary hardware on clustering while bringing forward low-cost, low-impact solutions to the user base. For example, we are doing interesting things with high reliability media mirroring over midrange and high speed data links to break the bond of the "shared SCSI" model of so many clustering solutions; this is a technology set that we are beginning to implement into standard clustering product offerings. It brings real value to the low-end NT Server space.

A secondary, but no less important, focus of this work is in expanding clustering beyond today's model of simply offering high availability with "fail over", and expanding more into the overall question of how to build a cluster that offers performance improvements without the user having to re-code their applications. This opens questions into automatic load balancing, process and thread migration, distributed I/O, and a whole host of other things that, while beat to death over the years, has never been adequately addressed in a manner friendly to the commercial server user. After all, finding economical solutions for the commercial user is the goal of all this (at least for NCR).

Steven McDowell
NCR Corporation
3325 Platt Springs Road
West Columbia, SC 29170
(803) 939-5091
steve.mcdowell@columbiaSC.ncr.com