2 AVIAT NETWORKS FEBRUARY 2011
RAN CAPACITY OPTIMIZATION THROUGH IPoTDM MICROWAVETRANSPORT
ML-PPP is defined in RFC 1990, and constructed as a PPP extension to allow for thecombining of multiple individual PPP links (dubbed “bonding”) into a single logical bundle.Each individual physical link is controlled by the PPP protocol, with ML-PPP residingbetween the data link and the network protocol layer and serving as the aggregation entity.As a result, multiple physical circuits or channels can be bonded to form one higher capacityvirtual circuit, and bandwidth can be flexibly scaled in E1/DS-1 increments.ML-PPP bundle membership negotiation offers an important tactical tool in that it can beused to change the aggregate capacity via addition or removal of individual PPP links and forpurposes of designing resilience into the network. Additionally, operators can manageconverged transport of legacy and adjunct IP traffic in a way that allows the allocated ML-PPP capacity to be adjusted as IP bandwidth demands rise. Lastly, ML-PPP can beemployed in conjunction with RFC 2686 multi-class extensions, providing the ability toprioritize traffic. Packets waiting to be transmitted over the ML-PPP tunnel are sortedbased on priority, with suspension of transmission of low-priority fragments until all high-priority fragments have been first serviced.
CONSIDERATIONS FOR IMPLEMENTATION OVER MICROWAVE
Once a multilink session has been successfully negotiated by the Link Control Protocol(LCP), ML-PPP can fragment incoming PDUs according to the available capacity andtransport these fragments in parallel across the physical links or circuits. On the receivingend, the fragments are properly sequenced based on their PPP headers, and reconstructedto the original PDU format. The latency is thus reduced to the amount of time required forthe last PDU fragment to reach the far side of the link, resulting in substantially decreasedlatency relative to the sequential transmission of PDUs over any of the individual links.The use of fragmentation, Ethernet framing overhead removal, and IPv4/v6 headercompression techniques as defined in RFC 2507 and RFC 2509 are crucial considerations formicrowave transport, both from the perspective of latency and capacity optimization. Asmall unfragmented PDU is encapsulated with a PPP header and a ML-PPP header,whereas when fragmentation is enabled, all but the first fragment require only the PPPheader for transmission. When properly optimized, operators can realize dramaticimprovements in effective microwave capacity of up to 40%.
As an example, a GSM codec may generate 33 Bytes of payload, encapsulated in 64 Byte Ethernetframes at the data link layer, and with the IPv4 header consuming 20 Bytes. As a result of theoverhead reduction achieved through the IPoTDM approach, a TDM microwave with an actualtransport capacity of 64 Mbps (32 x E1) can now carry the equivalent of approximately 90 Mbps ofuncompressed voice traffic load.
At the same time, the latency reduction achieved through fragmentation of longer PDUs willbenefit delay-sensitive voice and synchronization packets.If packet-based synchronization (Precision Time Protocol/IEEE 1588v2) is neither desirednor required by the NodeB, operators can continue to utilize the traditional E1/DS-1 linesynchronization method, as timing is maintained underneath the PPP/ML-PPP layers. As aresult, the IPoTDM architecture is not subject to synchronization discontinuities, and itsintroduction can be decoupled from the challenges in migrating to new synchronizationtechnologies.