Summary

Enterprise and SP networks are complex, interdependent entities. Enterprise network managers seek to improve business processes and workflow efficiencies by leveraging their technology. Service providers can help them achieve this by offering advanced managed or unmanaged (billable) services, such as VPNs. Both types of network have to be managed effectively using dedicated technology. We focus on SNMP-based network management, but it is important to note that this not the only approach. The trend in networking is towards what we refer to as aggregate objects . These can be seen in the many variants of interconnection technologies, such as VLANs. VLANs allow for LANs to be scaled upwards in a controlled fashion because the broadcast domain can be partitioned. This means that individual VLAN members (e.g., the software engineering department) can communicate within the one broadcast domain without its traffic crossing into a neighboring VLAN (e.g., the sales and marketing VLAN). Traffic crosses VLAN boundaries only as required, and this occurs using layer 3 routing. The mix of technologies involved in VLAN-based environments gives rise to aggregate objects. These objects in turn present scalability challenges to network management.

A successful NMS is one that maintains an accurate and up-to-date picture of the managed network. This is a lot harder than it sounds, particularly with the complex mix of technology and traffic types (many now have stringent real-time requirements) found in networks.

NMS constituent technology tends to follow a client/server architecture with many products based on Java technology. A typical NMS product offers a range of applications that fulfills the basic FCAPS areas as well as others, such as reporting and multiclient control.

SNMP provides a distributed model that uses managed-object schema definitions (MIBs) on remote devices. Instances of managed objects can be retrieved from agents on remote NEs. This can be done by a manager in conjunction with a local copy of the agent MIB; that is, there are two copies of the MIB. MIB structures often must be reflected in the data model (more on this later, but for now the data model is the way the NMS looks at the information relating to the managed objects). For this reason, the NMS quality can suffer if the MIBs are badly written. The mapping of MIBs to real NEs is reasonably easy to understand, particularly after using a MIB browser application (some are freely available on the Web).

A security scheme protects the agent data as well as the data in transit from the agent. A notification mechanism allows agents to asynchronously send messages to a manager when important events (such as faults) occur.

SNMPv3 offers a small number of protocol messages designed to allow effective management of NEs.