Enterasys Networks X-PeditionTM manual Multilink Header Option Format

PPP Features

MLPPP Packet Fragmentation and Serialization Transmission Latency

MLPPP’s packet transport method over multiple member links is made possible by fragmenting the packet after balancing the load bandwidth to fully utilize the member links’ bandwidth. When sent over a MLPPP link, each fragment carries a sequence number within the Multilink header, as shown in Figure 8-5, to ensure that fragment is reassembled and forwarded to higher layer applications in the same order.

Figure 8-5 Multilink Header Option Format

Type

Length

Code (Long/Short Sequence #)

#of Suspendable Classes

Additionally, each fragment of a sequence stream is assigned a class number in the MLPPP header, permitting at most four classes for the short and 16 for the long sequence number fragment. The higher the class number, the higher the priority it is granted over the line. For example, voice, video and data packets can be assigned high, medium and low priority sequence numbers to ensure proper QoS. Refer to “Configuring Quality of Service” on page 12-1for more information.

Since standard MLPPP allows only a single stream sequence number, the result is that only two priority level layers can be utilized without breaking the packet order as follows:

•Higher priority layer packets are sent without fragmentation and multilink headers.

•Lower priority layer packets are fragmented and interleaved with the higher priority packet. MLPPP is marked by the following limitations:

•A higher priority packet can be sent through only one member link since it is not fragmented and does not contain a sequence number, otherwise the higher priority packet order can not be guaranteed.

•The bandwidth of the higher priority packet should not exceed the speed of the designated link used. The result is MLPPP bundle bandwidth is not fully utilized.

Each MLPPP packet holding a fragment is transmitted through the member link with packet transmission latency occurring as the result of packet size operating against link speed. With a Multi-link PPP connection, most packets are fragmented into equal size fragments and transmitted over all member links to balance bandwidth loading over each link with the same or different speeds. To sum up, fragment size must be controlled in order to minimize transmission latency. Serialization transmission latency, measured in milliseconds, equals fragment size (in bits) multiplied by link speed (in Kbps) as shown in Table 8-1.

Table 8-1 Serialization Latency for Different Fragment Size/Link Speed

Fragment Size

Link	1 byte	64 bytes	128 bytes	256 bytes	512 bytes	1024 bytes	1500 bytes
Speed
56 kbps	143 us	9 ms	18 ms	36 ms	72 ms	144 ms	214 ms
64 kbps	125 us	8 ms	16 ms	32 ms	64 ms	126 ms	187 ms
128 kbps	62.5 us	4 ms	16 ms	32 ms	32 ms	64 ms	93 ms
256 kbps	31 us	2 ms	4 ms	8 ms	16 ms	32 ms	46 ms
512 kbps	15.6 us	1 ms	2 ms	4 ms	8 ms	16 ms	32 ms
768 kbps	10 us	640 us	1.28 ms	2.56 ms	5.12 ms	10.24 ms	15 ms

8-6 Configuring PPP