Elanti Network Optimization System

Intelligent routing for MPLS networks

The Elanti Network Optimization System (ENOS) is a patent pending real time traffic management system that optimizes traffic flows in IP MPLS networks. It provides service providers with three key capabilities:

1. Real time rerouting of IP traffic based on network failure or congestion scenarios, providing
   for a high level of service quality assurance.
   
   
2. Preventive optimization of traffic flows based on predictive analysis, preventing network congestions and reducing overall service outage times.
   
   
3. Prioritization of traffic classes based on service level definitions, enabling the assurance of different service levels leading to a more targeted service offering for the broader market.

The Elanti Network Optimization System is essential to allow service providers to begin offering high levels of service guarantees.

Benefits of Intelligent Routing

For many years the prevailing consensus among network operators has been that the only practical way to manage the performance of IP networks is to ensure that sufficient spare capacity is available in the network to deal with unforeseen traffic increases or network element failures. The lack of appropriate tools to manage traffic flows on IP networks and the acceptance that IP networks simply provided a best effort service for all users has contributed to this view. As more and more applications are IP based and move to the edge to the network (i.e. IP video conf., IPTV, IP collaboration, etc.), customers are demanding high levels of service guarantees in order to effectively manage their businesses. This forces service providers to look for new mechanisms to move from best effort to service guarantees.

With the advent of MPLS there is now a shift away from this simplistic approach. Over-provisioned networks are an uneconomical way of providing services in a competitive environment and now the emphasis is on maximizing revenue by providing different classes of service whereby customers pay according to the class of service they select.

Optimization of the traffic flows on IP/MPLS networks, even for the current single class of service environment, can be used to achieve very significant improvements in network efficiency and performance experienced by the customer, as demonstrated by the network simulation results below. In this example, the traffic demand before the onset of significant packet loss occurs can be increased by approximately 60% compared to routing traffic with standard shortest path routing.

Elanti Network Optimization System Architecture

The architecture of the Elanti Network Optimization System is shown below. The system consists of three main functional blocks: the Intelligent Route Selector, the Data Collector, and the Label Switched Path (LSP) Controller.

The Data Collection System periodically collects the actual network traffic levels from the underlying network elements. This data consists of the amount of traffic carried by each Label Switched Path (LSP). The traffic being carried by an LSP is measured by the router at the head of the LSP (the head-end router).

The Intelligent Route Selector performs a very fast global traffic optimization to determine the network paths which will provide the best network performance and the most effective use of the network.

The Label Switched Path Control System receives the required LSPs from the Intelligent Route Selector and generates the required control messages for the head-end routers in order to direct the traffic onto the required LSPs. The router control messages can use standard management protocols such as SNMP, or vendor specific mechanisms.

The complete Elanti Network Optimization System provides a comprehensive system for optimizing traffic flows which also includes modules for system performance reporting, network element alarming and network initialization and control.

The advantage of the Elanti Network Optimization System is that:

• The system has complete knowledge of all the traffic flowing in the network and can therefore make routing decisions which make best use of the network resources. This allows service providers to utilize better their network resources and therefore reduce their capital expenditures.
  
  
• The system uses the actual traffic distribution as input to its routing decisions which is more accurate than relying on theoretical traffic models or estimated traffic values. This allows service providers to meet service level contracts through real time traffic optimization for abnormal network behaviors.
  
  
• The system simplifies the process of traffic management by automating the complete tasks of data collection, path determination and configuration. This enables service providers to supplement traffic engineering with real time traffic optimization and reduce significantly their operating costs.

Intelligent Route Selector Technology – real time network control

The Elanti Network Optimization System’s Intelligent Route Selector technology combines the benefits of traditional mixed integer linear programming and heuristic methods in order to generate optimal or near optimal solutions in real time.

The Intelligent Route Selector uses a combination of machine learning and heuristic techniques to quickly identify a set of paths that can accommodate the current set of traffic flows. Because of its learning capability, the Intelligent Route Selector is able to adapt to new traffic and failure scenarios. It uses a Neural Network (NN), trained on a wide ranging set of traffic and failure scenarios. In operation, the NN quickly identifies a set of paths that form an approximate solution to accommodate the flows. Because of its generalization capability, the NN is able to identify a good approximate solution for scenarios which it has not seen during training. This approximate solution is then passed on to a Marginal Increase Heuristic (MIH) for further refinement to ensure that the final paths do in fact meet the requirements imposed by the flows. The purpose of the NN is to provide a starting point for the MIH which is as close to optimal as possible. This leaves the heuristic with minimal work to do in generating an optimal or near optimal solution. This combined NN/MIH system forms the heart of the Elanti Network Optimization System.

The Elanti Network Optimization System is aimed at real time network control leading to continuous optimization of traffic flows in the IP MPLS network. As such it differentiates itself from off-line systems for periodic network optimization or longer-term capacity planning functions, by virtue of its speed. This allows service providers  to guarantee much higher service qualities than conventional routing mechanisms. Given a measured set of traffic demands, the Elanti Network Optimization System can produce an optimal set of routes in less than one second on standard PC hardware.

Summary

The Elanti Network Optimization System provides a fast and accurate solution for routing traffic in MPLS networks in order to realise the best use of network infrastructure and the best network performance for customers. Routing network traffic flows according to an appropriate optimization objective means very substantial increases of greater than 60% can be made in the traffic carried by the network.