CD-adapco®, a leading computer-aided engineering enterprise, asked QLogic to create a solution that would provide shared storage access to multiple compute clusters, thus consolidating storage usage and providing faster, easier data access. QLogic responded with a two-phase approach that enabled CD-adapco to implement the new architecture at a minimal cost.
The combination of slow data copying and single points of failure increases the time required to balance the workload between the two clusters. For example, if one cluster is full and the other is empty, a user must move jobs to the other cluster to balance the work load. To move jobs, all files associated with the jobs must be copied from one storage cluster to another through the Ethernet network path. This process requires the user to manually move the data, and often results in storing two copies of the data. Not only is this an inefficient use of space, it also increases the chance that the data sets will not match.
Finally, there are non-redundant components in this solution: a single HP head, a single HP I/O node, a single IBM head, and a single IBM I/O node. If one of these components fails, its corresponding cluster and all the cluster's data would not be available until the failed component is replaced.