There has been a lot of discussion about whether Multi-protocol Label Switching is good or bad for the Internet. It's been called a social disease, something that should never be allowed on the 'Net, a disaster waiting to happen.
Strangely, none of this seems to have slowed its adoption - a majority of the service providers I've surveyed are either deploying MPLS or considering it.
So is MPLS bad for the 'Net?
To understand the answer, you have to understand something about the major objections. There are two:
First is that MPLS fundamentally breaks the Internet paradigm in two major ways. By supporting tunneling, it breaks the transparency paradigm. By supporting sessions, it breaks the datagram model. Both of these are fundamental architectural principles of the Internet.
The second major objection is that MPLS - like other connection-oriented technologies - doesn't scale infinitely. The Internet already has known scalability issues, which MPLS doesn't solve.
So if one takes a purist, Internet-centric approach, the naysayers are right: MPLS breaks some critical Internet architectural principles, while simultaneously failing to deliver any substantive incremental value, such as improving scalability.
But purists are missing some significant points. Specifically, MPLS was not designed to enhance the 'Net per se. Instead, it provides value to providers of IP and Internet services, including the following:
-- By building in support for quality of service and session-oriented services, MPLS lets providers of IP and Internet services better position those services for end customers, particularly large businesses.
-- Additionally, it lets service providers lower operating costs by providing an infrastructure that can consolidate IP, frame, ATM and other Layer 2 services.
-- Finally, MPLS IP VPNs set the stage for intercompany IP extranet communications.
On the issue of scalability, MPLS helps in a way that Internet purists might have overlooked. Most Internet purists might think of scalability as the ability to interconnect multiple networks effectively - and they're right, MPLS doesn't help here.
But the Internet has changed considerably over the past two years. Instead of hundreds to thousands of independent ISPs, the majority of 'Net traffic is now handled by a handful of large networks operated by general-purpose service providers (AT&T, WorldCom, British Telecom and others).
Providing a consolidated infrastructure for these large providers (at least in theory) can reduce their network operational and management costs - which represent the largest single component. Anything that reduces these costs helps scalability.
The bottom line is that MPLS might not help the 'Net, but it helps 'Net providers. And that's why it's here to stay.