Autonomous Semantic Boundary Aware Network Topology Discovery of Large, Multi-Subnet Networks

Abstract

Active Network Topology discovery is a mechanism to keep track of the status of network resources and their interconnections. The knowledge of the real-time topology of a network is crucial for management tasks such as resource management, resource discovery, congestion avoidance, failure detection and correction as well as network state visualization and analysis. In today’s networks ensuring scalability by manual entry of topology information is very difficult due to their size and dynamic nature. Therefore we need to define mechanisms to automatically discover the network topology. Also sending request to all possible IP addresses in a network is not feasible because of the large number of possible IP addresses e.g. there can be more than sixteen million possible addresses for a Class A network. Therefore in this paper we propose an efficient algorithm for the automatic discovery of the network resources where we use ARP cache based guessing to check the subnets with high probability of being alive. For status checking of network nodes current techniques use ICMP based probing, which are be blocked or filtered by many routers. Therefore we use an open source TCP based ping mechanism (TPing) [1] for discovery of network elements. We make use of concurrent TPing to discover the topology in multi-subnet environments and our algorithm performs discovery of networks employing CIDR (Classless Inter- Domain Routing) based IP addressing scheme efficiently and accurately. TPing is only available for Linux and for platform independence I developed my own windows version of that. Our algorithm also discovers the MAC addresses with multiple hop distance which are transparent from subnet to subnet. The algorithm only requires SNMP (Simple Network management protocol) on routers and switches and do not need SNMP on end hosts for the discovery process. The data is exported in a semantically enriched form. By defining shared and common vocabularies, ontologies help both people and machines to communicate concisely, supporting the exchange of semantics instead of syntax. We are using Resource Description Framework (RDF) to export the data from the discovery algorithm. This approach will reduce the human intervention for the information retrieval about a particular network so that it can be further used by FIPA compliant Multi-agent systems to provide an agent understandable format where agents will be acting on behalf of the user for automatically discovering and querying the topology ontologies and retrieving the required data.