are increasingly using the Internet as form of WAN transport, therefore QoS
models needs to be revisited. QoS works effectively when deployed in an
service-level agreement (SLA) environment today, like Multiprotocol Label
Switching (MPLS) . The available bandwidth on the internet at a given point of
time can vary, and can be often much lesser than the actual bandwidth offered
by the service provider. In cases of non SLA environments, QoS has limitations
- mainly because it cannot predict changing bandwidth on the link.
WAN (IWAN) recommends using Dynamic Multipoint VPN (DMVPN) over Internet to
connect branches to the data center or headquarters, and QoS to be deployed in
such environments of fluctuating bandwidth. Currently, the shapers that are
applied as part of the egress QoS policy are static in value - they are
configured based on the service provider bandwidth offering, they do not change
with time and hence do not reflect the actual available Internet bandwidth. In
many instances where Internet available bandwidth becomes much lesser than the
offered bandwidth, the shapers become irrelevant as they do not adapt to the
varying bandwidth. Due to the static value of the shapers, application traffic
gets dropped indiscriminately at the Internet core, nullifying the very need to
have configured a QoS policy to protect critical traffic.
DMVPN provides the
ability to do QoS per-tunnel, which means a QoS policy can be applied at the
hub towards a specific spoke, to ensure a high bandwidth hub does not overrun a
low capacity spoke. However, these QoS policies still work with static shapers
per spoke. If the bandwidth towards a particular spoke fluctuates, the shapers
towards the spokes do not adapt. Also, it is not possible today to configure a
QoS policy for the traffic from the spoke towards the hub, which is very common
in many retail-like environments.
The Adaptive QoS
over DMVPN feature provides the following benefits:
shaper parameters based on the actual available Internet bandwidth in both
directions that is periodically computed.
configure a QoS policy on the spoke towards the hub.
control of application performance at the enterprise edge even in changing
bandwidth scenarios over the Internet.
tunnel shape adaptation to provide effective bandwidth between spoke and hub.