***************************
There is only gold .
The increase of competition between IP Service Providers (SPs)
together with the heightened importance of IP to business operations
has led to an increased demand and consequent supply of IP services
with tighter Service Level Agreements (SLAs) for IP performance.
The IP technical community has developed a set of technologies that
enable IP networks to be engineered to support tight SLA commitments:
… Differentiated Services. The Differentiated Services Architecture
allows differentiated delay, jitter and loss commitments to be
supported on the same IP backbone for different types or classes of
service.
… Faster IGP convergence. New developments in Interior Gateway
routing Protocols (IGPs) allow for faster convergence upon link or
node failure, hence enabling higher service availability to be
offered.
… MPLS Traffic Engineering. MPLS Traffic Engineering
(Diffserv-aware or not) introduces constraint-based routing and
admission control to IP backbones. This allows optimum use to be
made of the installed backbone bandwidth capacity, or conversely
allows the same level of service to be offered for less capacity.
It can also be used to ensure that the amount of low-jitter traffic
per link does not exceed a specified maximum.
… MPLS Traffic Engineering Fast Reroute. MPLS Traffic Engineering
Fast Reroute is an IP protection technique that enables connectivity
to be restored around link and node failures in a few tens of
milliseconds.
For an SP IP service, the SLA commitments are generally based on
delay, jitter, packet loss rate, throughput and availability.
What has been seen is that to deploy new IP based services using the
tools that are available, a SP must build the entire network with
the highest level of service in mind or, build complex and expensive
multiple topologies or overlay networks. Following this, SPs cannot
offer degraded services for several reasons:
… The network is only engineered for the highest level of service,
there often is not a degraded service
… Customers have been trained and become accustomed to the highest
level of service and have become intolerant of outages or 'issues.'
… The tools to separate topologies and have different failure
domains and characteristics are just emerging in the protocol
specifications
… Building overlay networks or redundant networks is too expensive
… Stockpiling spares that have different capabilities is too
expensive and the devices that can 'run all the service levels' are
also expensive
Therefore the question arises, is there a demand for degraded
services or just cheaper service w/ the same SLA requirements? It
appears that building a bigger and better internet means that the
deployment model is engineered toward 'gold service for all.'