Providing Open Architecture High Availability Solutions

Providing Open Architecture High Availability Solutions
96
In addition, service group dependencies also must be modeled, monitored and managed. Service
dependency groups are collections of managed components orchestrated to cooperate in
redundancy configurations in order to secure the availability of the services the system provides.
These sometimes-complex dependencies and interdependencies not only overlap architectural
layers, but they often involve more than one node and include heterogeneous components and
platforms.
The topology and data represented in the model must be stored in a high-performance database in
order to provide current state data that can be rapidly updated to reflect ongoing changes of state in
the managed components. In addition, to secure the availability of the management service itself,
the database should ideally be replicated to a redundant node in case the component hosting the
primary management database should fail.
Managing component faults and failures requires rapid access to the schema of the physical
(parent/child) dependencies that enable the system to function. Understanding and accessing the
knowledge in this system model quickly is crucial to effective service availability management.
10.3 Checkpointing Data to Redundant Components
Redundancy is key to availability. In many cases, components are stateless and can be replaced on-
the-fly by a redundant standby component without interrupting service. However, in some cases
the component contains state information that must be preserved to avoid interrupting service
during switchover to a standby component.
Figure 17. Example Directed Graph Describing a Managed System
A8680-01
Apps
App
2
App
1
Power
PS
2
PS
1
PS
3
Fans
Fans
2
Fans
1
Fans
3
Host H/W
2
Host H/W
1
O/S
2
O/S
1
Comms
App
4
App
3
Service
1
Line H/W
4
Line H/W
3
O/S
4
O/S
3
App
6
App
5
Line H/W
6
Line H/W
5
O/S
6
O/S
5
Directed Graph
Group Membership
Dependency