To enable high availability for SAP business scenarios, you must set up your networks so that there is no single point of failure. For a server network, you need to secure the following:
· Network interface of the server
· Network itself
One possibility to improve availability is to set up redundant network access for a server, using redundant network cards. In the event of network failure, a switchover is performed to the redundant card, including transfer of the IP address (alternatively, it is possible to address several network cards with the same IP address). Such solutions are partially incorporated in the operating system, are implemented as cluster products, or can be controlled by scripts that you write yourself.
The redundant network cards ought to contain connections to different network segments, so that they can survive the failure of an active network component.
However, the use of redundant network cards means you need identity or IP address takeover solutions, in case the server needs to be reached by an alternative network path. Cluster solutions for servers generally offer the possibility of addressing using a virtual IP address, independently of the server’s internal IP address.
Note the following general rules:
· A separate LAN segment for the server local area network (LAN)
Due to performance, administrative, and security reasons, use a separate LAN segment for the server network. This enables more efficient administration and improved high availability.
· Setting up the hosts close together
Connect your hosts directly to the server network rather than across a decentralized network.
The graphic below shows a server farm using the currently popular method of switching technology, which guarantees the redundancy required for high availability in the server network:
Redundant Server Network
Note the following features of this solution:
· Redundant active network components
This ensures that, if an active network component fails, the redundantly connected servers can still access the network.
· Redundant network adapters in the database and central instance hosts
This ensures that the:
Ў Redundancy of the physical network can be used
Ў Network adapter is not a single point of failure
Switchover software is used to switch between the network adapters and between the protected servers.
· Critical servers connected to multiple physical sub-networks
This ensures that, even if a physical network segment fails, the servers can still access the network using an alternative network segment.
· Application servers that are not redundantly connected are at least connected to different physical network segments
This ensures that, if a physical network segment fails, access is not lost simultaneously for all application servers. However, the application servers connected to the failed network segment can no longer access the network, and are therefore unavailable.
For more information about the zSeries specific recommendations on how to set up a high availability network between the ICLI client and server see the IBM documentation High Availability for SAP on zSeries Using Autonomic Computing Technologies (SC33-8206).