HP StorageWorks 2/24 FW 07.00.00/HAFM SW 08.06.00 McDATA Products in a SAN Env - Page 123

Fabric Availability Planning Considerations for Fibre Channel Topologies 3 The fabric is divided into four performance regions as follows: • Local Tier 1 devices - A video server application with I/O capabilities of 40 MBps and 2,000 IOPS must be connected to the fabric. Because the application is critical and high bandwidth (in excess of 35 MBps), the server and associated storage are directly attached to the core director as Tier 1 devices. No ISLs are used for server-to-storage connectivity. • 11 to 1 fan-out region - Eleven NT servers with I/O capabilities of 10 MBps and 1,000 IOPS are fabric-attached through a 32-port edge switch. The primary applications are e-mail and online transaction processing (OLTP). Because bandwidth use is light and noncritical, the servers are connected to the core director with a single ISL that is intentionally oversubscribed (1.1 Gbps plus Class F traffic). The servers are connected to storage devices with I/O capabilities of 11,000 IOPS. • 6 to 1 fan-out region - Six servers with I/O capabilities of 20 MBps and 1,500 IOPS are fabric-attached through a 16-port edge switch. Bandwidth use is light to medium but critical, so the servers are connected to the core director with two ISLs (0.6 Gbps each plus Class F traffic). The servers are connected to storage devices with I/O capabilities of 9,000 IOPS. • 3 to 1 fan-out region - Three servers with I/O capabilities of 30 MBps and 2,000 IOPS are fabric-attached through a 16-port edge switch. Bandwidth use is medium but non critical, so the servers are connected to the core director with one ISL (0.9 Gbps plus Class F traffic). The servers are connected to storage devices with I/O capabilities of 6,000 IOPS. Many fabric-attached devices require highly-available connectivity to support applications such as disk mirroring, server clustering, or business continuance operations. High availability is accomplished by deploying a resilient fabric topology or redundant fabrics. A fabric topology that provides at least two internal routes between fabric elements is considered resilient. A single director, switch, or ISL failure does not affect the remaining elements and the overall fabric remains operational. However, unforeseen events such as human error, software failure, or disaster can cause the failure of a single resilient fabric. Using redundant fabrics (with resiliency) mitigates these effects and significantly increases fabric availability. Planning Considerations for Fibre Channel Topologies 3-37