Figure 3. Two-data center Extended Cluster for RAC—up to 100 km currently-supported configuration

Test architecture

The test configuration for the 100-km evaluation followed the supported configuration for all subsequent SGeRAC implementations.

Supported hardware configuration

•Two-node cluster (PA-RISC or HP Integrity servers, not mixed).

–A two-, four-, six- or eight-node cluster is supported up to 10 km with VERITAS Cluster Volume Manager (CVM).

•Redundant heartbeat, plus one or two LANs for Cache Fusion traffic.

–Clusters with VERITAS CVM; only single-heartbeat subnet is supported (use primary/standby LANs).

•Any Fibre Channel storage (using HP StorageWorks Virtual Array (VA), HP StorageWorks XP disk array, HP StorageWorks Enterprise Virtual Array (EVA), or just a bunch of disks (JBOD) technologies).

–Note: EVA disk arrays do not currently support VERITAS Volume Manager (VxVM)/CVM.

•Both TCP/IP networking and Fibre Channel data can go through the same DWDM box; redundant DWDM boxes in each data center are not required if the box is designated as acceptably fault- tolerant.

•At least two dark fiber links between data centers.

–During testing, AT&T-supplied dark fiber optic links were used between data centers (to further enhance disaster tolerance capabilities, alternate physical routes can be added to remove single points of failure).

–Note: TCP/IP networking and Fibre Channel data pass through the same DWDM box; in a non- test environment, redundant DWDM boxes in each data center are not required if fault-tolerant design specifications meet disaster tolerance requirements.

9