Understanding and Designing Serviceguard Disaster Recovery Architectures

There can be up to 16 nodes in this cluster (this includes the nodes in the Arbitrator data center,

unless they are running the Serviceguard Quorum Server).

You can use SONET instead of DWDM in this configuration by replacing the DWDM boxes with

SONET boxes. With SONET, you can use a point to point topology, as is shown in this example,

or you can use a dual SONET ring topology which passes through all three data centers, where

the working ring and the protection ring are alternately routed.

Figure 28 Extended Cluster for RAC with Two Data Centers and a Third Location with WDM Links

Quorum server - Third Location

Primary Data Center 2Primary Data Center 1

QS Node

FC Switch

This network is

used for RAC

cache

fusion trafﬁc.

Node A

Node B

Node C Node D

Maximum distance

100 kilometers

DWDM

FC Switch

Network Switch 1 Network Switch 5

Network

Switch 4

Network Switch 2

Network

Switch 3

FC Switch

Storage Storage

Redundant DWDM

DWDM

WAN

Network Switch 6

This example, figure 29 is specifically for an EC RAC cluster. The WDM boxes connected between

the two Primary Data Centers are configured with redundant fiber links and the standby fiber

feature is enabled. Note that there is a separate network (indicated by the green lines to switches

3 and 4) being used for the RAC Cache Fusion traffic to ensure good RAC performance. Switches

2 and 5 are used for the Standby network, which can provide local LAN failover for both the

Primary Heartbeat network and the Primary RAC Cache Fusion network. However, the Standby

network can only provide local failover capability for one of the Primary networks at a time.

Therefore, it is preferable to have a separate Standby network for the Heartbeat network and for

the RAC Cache Fusion network. The distance between the Quorum Server Data center and the

Primary Data Centers is not specified, however you must ensure that the Quorum Server can be

contacted within a reasonable amount of time (should be within the NODE_TIMEOUT period for

Extended Distance Cluster on HP-UX 67