Networking: MPLS or the Internet? Not an All or Nothing Deal

By Jim Metzler

July 1, 2008 —  (Page 1 of 3)

Having spent previous columns discussing MPLS (Multi-Protocol Label Switching) technology, I now want to introduce to the discussion a different approach to wide area networking—an approach based on leveraging the Internet. This is not an all-or-nothing proposition. I do not see the question facing IT organizations as being, “Do you use MPLS or do you use the Internet?” Rather, I see the question as being, “When and where do you use MPLS and when and where do you use the Internet?”

Some reasons why MPLS has been so successful in the marketplace are that it provides predictable latency and packet loss characteristics along with quality-of-service functionality that allows for the convergence of voice, video and data traffic within a single service offering. However, one of the disadvantages of MPLS is that, in most instances, it is relatively expensive and there tends to be a long lead time associated with deploying new MPLS services.

MPLS also tends to be complex, precluding virtually all IT organizations from deploying private MPLS networks. Instead, when an IT organization acquires MPLS services from a carrier, (e.g., AT&T) they usually acquire a managed service.

As is the case with most services provided by a carrier, MPLS services typically come with a service level agreement (SLA) that describes the key characteristics of the service. This can include time to install a new service along with metrics for the promised availability, delay and loss. If the carrier does not live up to the SLA, there are penalties.

The problem is that the SLAs associated with MPLS services tend to be weak. For example, the SLAs are primarily reactive in focus; e.g., the computation of an outage begins only when the customer opens a trouble ticket. Not only is the computation of the SLA metrics done in a way that is unfavorable to the customer, the level of compensation for violation of SLAs remains quite modest.

Further weakening the value of these SLAs is the fact that the SLA metrics are primarily calculated as network-wide averages rather than for a specific customer’s traffic or for a particular site. As a result, it would be possible for a company’s data center to receive notably poor service in spite of the fact that the network-wide SLA metrics remain within agreed bounds. In addition, many of the carriers have some unique quirks in their SLAs. For example, one carrier excludes from their availability target any network outage of less than a minute in duration.

Related Search Term(s): Networking, Akamai

Share this link: http://www.sysmannews.com/link/32468