Service Assurance: What Happens in Pakistan Stays in Pakistan (Mu Dynamics blog)
by Thomas Maufer on 28 February 2008 - 04:36:26 PM
Service Assurance for Next-Generation Networks is a Layer 2-7 Problem
(Part I - Mistakes Happen)
Providers are actively rolling out next Gen IP services using IPTV and IMS to bolster revenue. For IMS, VoIP
is a key application, though far from the only application. All of
these services have a shared dependence on high network reliability,
availability, and security. In many cases (such as, for example,
AT&T's deployment of IMS, which is just one "application" on their purpose-built converged network backbone, based on MPLS), the network infrastructure is vertically integrated, all the way down to layer-2 (MPLS)
and all the way up to layer-7, to ensure that the network can deliver
Quality-of-Service suitable to the application(s) being delivered.
Quality of Service is a "term of art" in the networking business. Very narrowly, QoS
defines priority queuing in switches and routers and higher-layer
gateways that deliver a data stream across the network with predictable
latency and jitter. A more important metric is "Quality of
Experience," applying to the paying customer's perception of the
network's ability to deliver a usable service. When defined narrowly,
"quality of service" is relatively easy to measure, but no amount of
queue management can compensate for a network that is unstable due to
poor software implementations of the many protocols, perhaps dozens of
them, that must work together seamlessly to deliver a service.
Quality
and Service really can't be separated when revenue is at stake. At the
end of a billing cycle, a service lacking quality -- regardless of
measurement units -- will not be able to retain customers. Churn is
very expensive for providers: Losing one subscriber equates to the cost
of keeping as many as 10 subscribers since the cost of acquiring a replacement is
very high. In light of the big picture, quality of service clearly means
much more than enabling diffserv or IEEE 802.1p at layer-3 or layer-2, respectively.
But, what does this have to do with Pakistan? A significant IPTV
outage happened this past weekend because the Pakistan
Telecommunications Agency (PTA) operations group changed its router
configuration and told the world (perhaps by accident, perhaps not)
that they had an excellent route to part of the IP address space owned by YouTube. Of course, when the deluge of traffic arrived in Pakistan, it was dropped. With YouTube's "IPTV"
service was interrupted due to an external event, clearly the Quality of
Service was reduced or nonexistent. Granted, this was a semi-intentional action: The PTA
claims that this route advertisement was never supposed to leak out of
Pakistan. This is probably the truth as building global networks is a very
complex business and accidents happen. Routing in the Internet is all about policy, and policies change and frequently do not mesh perfectly. Because of the efforts of a
well-coordinated set of network engineers, who work together with their
peers at other providers (via forums like NANOG and other such organizations around the world), the Internet mostly works.
(to be continued on the next blog, "Bugs Happen.")
Comments:
Write a comment
- Required fields are marked with *.
|
