Multi Protocol Label Switching (MPLS), Frame Relay and ATM are packet-switching technologies that allow sharing of physical links in the service provider network among multiple customers using statistical multiplexing.In these environments, each switching device within the network cloud represents a congestion point where IP packets will be dropped or re-ordered as a result of traffic bursts and oversubscription, creating contention for network resources. This is particularly true at the edge of the service provider network, where LAN traffic is stepped down to smaller WAN bandwidth rates.
Network congestion typically leads to packets being dropped. While packet loss rates vary widely, it is common to see 0.05% to 0.1% packet loss in normal traffic hours across service provider networks, with peaks reaching 2%. With congestion comes jitter and latency, which has a major impact on real time applications, like VoIP and streaming media. Applications that leverage TCP are also affected by congestion, because TCP has a built-in backoff mechanism that throttles back network traffic during periods of congestion. The result is degradation in performance for many applications.
Silver Peak Solution
The following relevant Silver Peak features mitigate the impact of impairments that arise specifically in shared packet-switching networks:
Network Memory is used in conjunction with advanced compression techniques to reduce the amount of data traversing the WAN. This increase WAN efficiency, which helps to avoid congestion when possible.
Forward Error Correction (“FEC”) protects traffic from collateral congestion in a shared service provider network. The Silver Peak solution adapts FEC to accommodate changing WAN conditions, ensuring an optimal blend between FEC overhead and application performance.
QoS is used to prioritizes traffic types, and allocate bandwidth accordingly. This improves jitter and latency by enforcing specific policies that govern how to route downstream traffic.
Silver Peak can both honor and set DSCP/MPLS markings on a per-application basis for added QoS granularity.
