self configuration of networks


Overview - Description




Traditional configuration management involves complex labor-intensive processes performed by experts. The configuration tasks such as installing or reconfiguring a system, provisioning network services and allocating resources typically involve a large number of activities involving multiple network elements. The network elements may be associated with proprietary configuration management instrumentation and may also be spread across heterogeneous network domains thereby increasing the complexity of configuration management.

The traditional configuration management methods are based on the centralized manager/agent model. This approach involves centralization of configuration intelligence in a central management station like the SNMP Manager or the Policy Decision Point (PDP). However, an increase in the number and complexity of network elements can increase the complexity and overload the central management station. Besides, the central management station may have limited access to multiple heterogeneous management domains.

The purpose of this project has been to introduce architectures for the self-configuration and self-management of networks. The fundamental objective of the research endeavors associated with this project has been to associate configuration intelligence with the components of the network, rather than limit it to a centralized management station. The association of configuration intelligence with the network elements tries to overcome the drawbacks associated with the traditional manager-agent approach which is practiced in SNMP, CLI and Policy-based networking.




The various research endeavors associated with this project are described as follows:

  1. SELFCON An Architecture for Self-Configuration of Networks

The manual processes presently involved with network management are quickly reaching their limits as networks become more complex and implement emerging services. SELFCON is an architecture that combines several techniques such as object modeling using the standard DEN specification, directory services, and network programming to provide self-configuration of networks. This feature of self-configuration addresses several configuration issues encountered in the most commonly used configuration methods and protocols, such as CLI, SNMP and COPS.

The configuration information is maintained in a standard directory. The configuration information maintained in the directory is modeled using the standard DEN specification. The network elements register for change notification at the directory server using a directory service. The directory server notifies the network elements about any related changes in the configuration information. The network elements upon receiving the changed configuration data are able to perform self-configuration.

SELFCON allows increased scalability, as the network elements are able to automatically adapt themselves to the new configuration policies. SELFCON is able to eliminate configuration inconsistencies, reduce configuration management tasks, and is able to provide a unifying framework for effective enterprise management.


Policy-based networking (PBN) has emerged as a promising paradigm for network operation and management. It is based on high-level control/management policies, that is, rules that describe the desired behavior of the network in a way as independent as possible of network devices and topology. Two basic entities are distinguished: the policy enforcement points (PEPs) and policy decision points (PDPs). The PEPs typically reside on the managed devices. Their role is to enforce the commands they receive in the form of configuration data from the PDPs. The PDPs process the high-level policies along with other data such as network state information and generate configuration data for the PEPs.

The COPS protocol is designed to communicate self-identifying policy-related information, exchanged between the PDP and the PEP. In COPS, each PEP may support one or more clients of different client types; different client types exist for the different policing areas (security, QoS, admission control, accounting, etc). COPS for Policy Provisioning (COPS-PR) is one of those client types.

COPS-PR uses special structures called policy information bases (PIBs) that store policy information at the PEPs and control the behavior of the devices. However, the rigidity of its mechanisms constrains the intelligence that can be pushed toward the managed devices.

This rigidity can be relaxed by using meta-policies, rules that enforce the appropriate policies on the devices. Meta-policies are stored and processed by the devices, independent of their semantics, thus making the model more efficient, scalable, distributed, and robust.

The additional functionality is implemented through a PIB that stores and handles meta-policies. This way, although the philosophy of the conveyed policing information is now different, no modifications are required to the COPSPR protocol.

  1. Framework for Economical MPLS Protection

The growth of MPLS as the emerging choice for provisioning and managing core networks has placed significant emphasis on MPLS based recovery mechanisms. MPLS based protection can be categorized as Local protection or Global protection. In case of local protection, the switch over to the backup path is performed by the MPLS router (LSR) that is immediately upstream of the point of failure. In case of global protection, this switch over is performed by a pre-determined LSR known as Path Switch LSR (PSL). The PSL is usually distant from the point of failure.

The present global protection frameworks require extensive involvement of LSRs in the protection framework. The LSRs are involved in failure propagation and maintenance of information regarding upstream neighbors. This extensive involvement of the LSRs in the protection frameworks may become a scalability issue particularly in case of complex MPLS networks as a single failure may affect many LSPs and PSLs. The restoration time in the present frameworks is dependant on the length of the working path. This dependence leads to an increase in the number of PSLs required for effective protection, which thereby increases resource usage. The present frameworks also do not monitor the PSL or the backup path for fault notification - a feature that is important to ensure reliable protection frameworks.

We have proposed a framework for economical global protection in MPLS networks. This framework is the first framework that reduces the extensive involvement of intermediate LSRs in the protection process. The framework provides resource-effective and fast fault notification. The proposed framework also reduces the number of PSLs required for effective path protection. The substantial reduction in the involvement of intermediate LSRs and the reduction in the number of PSLs provides an economical approach for protection in MPLS networks. We are also able to monitor the PSLs, PMLs and the backup path in a cost-effective manner.