Optimization for High-Speed Network

When you e-mail someone, the addressee receives your communiqué in the exact form and format that you sent it in. Between your clicking the "send button and the recipient seeing it, however, the mail goes through a complex set of processes. The electronic information that underlies your mail is divided into several parts called "packets".

These packets are then pushed through a multitude of communication channels selected from the millions that constitute the Internet and delivered at the destination. Each of the packets might take a different route over the internet. The packets are then reassembled at the destination and presented in the exact same form in which you sent it.

The number of people using information technology is growing in leaps and bounds. At the same time, the devices that form the interface between the internet and its users are growing in power and sophistication. For instance, the average smartphone available today packs more processing punch than even desktop computers of just a few years ago. As a result, the underlying communication infrastructure that forms the "net" in the internet stands the risk of being overwhelmed by the growing volume of traffic it has to handle.

Network service providers, therefore, are in the unenviable position of having to run hard just to stay in the same place. Staying ahead of the curve of growing traffic requires constant investment in new equipment and new technologies. Apart from the capital expenses, operational expenses, too, keep increasing. The problem is compounded by obsolescence.

The standards and protocols that define the "dos and don'ts" to create the framework for information transmission in the digital world were created many years ago. While they worked just fine at the time, the growth of traffic on the 'net has been far greater than what the creators of these standards envisaged. This has resulted in bottlenecks which slow things down in spite of constant enhancements to the communication network.

One example of this is the Internet Protocol (version 4) address in its current format. These addresses are essential elements of the "from" and the "to" addresses necessary to identify each packet that is transmitted over the internet. The first part of the address identifies the network from which the data packet originates. Since this is defined by a preset number of characters, the total number of unique identities available at a given instant is limited too. This means that the system can only identify and forward so many packets at the same time, even though the communication channel might have been able to carry more packets.

In Ethernet environments, for instance, data packets, known as frames, are forwarded with unique identification codes known as Backbone VLAN Identifiers or BVIDs. Only 12 bits of the frame are assigned for the BVID, which means that only 4096 unique service instances can be supported simultaneously by the network.

One way to clear this bottleneck would be to reduce the dependence on unique network identifiers, so that more data can be processed and forwarded within the same time frame using existing hardware. An effective way to do this would give network service providers a means to enhance their network capacity without additional capital or operational expenditure.

At the IITB-Monash Research Academy, research scholar Deval Bhamare has been working to find a solution to this problem. Guided by IITB and Professor Mohan Krishnamoorthy of Monash University, Deval has been part of a team that has devised an elegant solution. Using advances in Carrier Ethernet technology, a new routing paradigm has been created.

Called Omnipresent Ethernet, the technology has been demonstrated on a test bed. It uses Centralized Control plane approach where the Network Management System performs the task of service provisioning, router configuration and connectivity fault management etc. It reduces the complexity and eventually cost of the router.

Since the route to be taken by each packet is already defined, the router is no role in the decision-making process. Therefore, it no longer needs to identify process or track the packets that are sent. This facilitates communication without large lookup tables. This means that the number of Ethernet frames that can be processed at a given instance is limited only by the capacity of the infrastructure. The limitations imposed by the constraints of the IP identifier are removed from the equation. The result is a faster, more effective, energy-efficient network that costs less in relation to its capacity. This collapses the layer 2 and 3 of the network into a single layer, called layer 2.5, eventually reducing the network costs.

Deval is confident that the work done at IITB-Monash Research Academy will touch the lives of the common man. "This new technology will help reduce the cost of adding new capacity to the broadband infrastructure. As a result, the cost of using broadband will come down, which in turn means that a greater percentage of the population will have access to IT."

Research scholar: Deval Arunkumar Bhamare, IITB-Monash Research Academy

Project title: Optimization for High-Speed Networks and Control Plane Architecture

Supervisors: Prof. Mohan Krishnamoorthy

Contact details: devalb@gmail.com

For more information and details on this technology, email research@iitbmonash.org