Internet of Tomorrow: MPLS
Date: Sunday , October 27, 2002
INTERNET TRAFFIC IS INCREASING EVERYDAY. And with increasing traffic comes increasing network demand. With growing demand for voice, data, and video carried over the Internet and each traffic type having its own characteristics, there are some difficult challenges facing the Internet.
Traditionally, internet traffic routing (or forwarding) is done based on a protocol call Internet Protocol (IP). Although IP has been an extremely successful and widely-deployed technology, it is too slow to handle large traffic loads and has bottlenecks when it comes to keeping up with the variety of traffic types and the demands placed on it. When IP is pushed to its limits, the network suffers from lost traffic and lost connections resulting in an overall poor performance.
A new protocol called Multi-Protocol Label Switching (MPLS) is gaining popularity to address the issues and concerns with IP routing. MPLS is a very simple and elegant method based on the idea of label-switching. It is designed to handle growing traffic and a wide variety of traffic types. MPLS is not meant to replace, but to complement, IP in the networks of today and tomorrow.
So what is this new idea of label-switching? The idea is analogous to the use of ZIP codes in the U.S. postal system. When we mail a letter to John Smith, 123 Some street, Some City-State, 99999, the postal network does not process the entire address. Just the ZIP code 99999 is used by the postal network to figure out where the letter is destined. Once it reaches the ZIP area 99999, the address (123 Some street) is used to forward the letter to the final destination. To compare this with the world of internet, the letter is analogous to a packet of data and the street address (123 Some street) is analogous to the IP address in that packet.
The IP forwarding makes use of the IP address inside of a packet to make the forwarding decision. The ZIP code in the postal example is analogous to the label that MPLS uses to forward a whole bunch of packets to the same destination without even looking at its IP address. Once this bunch of packets reaches the end-path designated by the label, the IP address is used to deliver the packets to the end-users.
In the past, the Internet has not been used for different types of traffic such as voice and video. And since the process of allocating resources for a given traffic type is fairly complex, a simplified approach was adopted in IP forwarding where all traffic was treated the same and it was delivered on a best-effort basis. If there was a packet loss, there was no mechanism in IP to restore it. Various methods of feedback (for example, congestion control) have been used to attempt to overcome this problem.
The IP forwarding along with these methods has proven to be adequate so far because the majority of traffic on the Internet has been data. However, voice and video have characteristics that are completely different from those of data traffic. Voice and video traffic, unlike data traffic cannot tolerate delay. In our phone conversations over today's telephony networks (PSTN), we are used to practically no delay. In some overseas calls, we do experience a little more delay but that is constantly being improved in our telephone networks around the world. Now, if we start carrying voice over the internet, we would expect the same quality of service (QOS).
While it may be acceptable that your email (data traffic) reaches your friend an hour after you send it, it would be silly if your "hello" (voice traffic) reaches your friend an hour after you say it. In other words, voice traffic needs to be carried over the Internet with very little (practically zero) delay. Moreover, there should not be any variation in the amount of delay (although small) introduced by the network. Clearly, it cannot be accomplished using the best-effort approach taken by IP forwarding.
MPLS is designed to solve this problem of delay. Since label-switching speeds up the process of traffic delivery, it reduces the delay significantly. MPLS would improve the quality of service for voice and video traffic significantly as long as the underlying network has enough bandwidth to carry such traffic. If the performance is poor due to low bandwidth network, MPLS will not improve the situation.
MPLS is playing an important role in the area of Traffic Engineering (TE). Carriers need to engineer their customers' traffic in such a way that it meets their QOS needs within the constraints of their network performance. TE focuses on the measurement and control of traffic. Traffic is organized as service classes A, B, C, or D with each of these classes having certain well-defined features.
A user's traffic would fit into one of these classes based on the traffic type and the user's QOS need. The Label Switch Routers (LSRs), which support the MPLS protocol, route this engineered traffic based on algorithms for these traffic classes.
Another application of MPLS is in the area of Virtual Private Networks (VPN). VPN is a geographically diverse network owned by an enterprise where access and security policies are the same as in a private network of an enterprise. The concept of VPN itself is fairly old, but MPLS is enabling a more reliable, scalable, and cost-effective implementation of VPNs. The MPLS/VPN implementation provides the capability to create private network services over a public infrastructure. It provides the dual benefit of simpler routing of customers' VPN traffic as well as simpler provisioning from the service provider's point-of-view.
While MPLS technology holds great promise, it is still in the early stages of adoption. It is about two years into a three- to five-year initial implementation phase. As such, the technology has not yet established a history of business success. This leads to service providers dragging their feet in justifying a major overhaul of their network infrastructures and the back-end support systems that accompany them.
At the same time, service providers don't want to risk being late to market with new-generation IP services because of a poor decision to postpone new infrastructure investments. To minimize risk while maximizing revenue, service providers should consider an MPLS migration path that enables them to move in stages to telephony-grade IP/MPLS network infrastructures.
With MPLS technology in their hands, service providers will be ready to deliver it to their customers as soon as it becomes a competitive necessity. This phased migration must be conducted in such a way that their existing customers are not impacted and their current revenue stream is not jeopardized.