Enterasys Networks X-PeditionTM manual Fragment Interleaving Over the Link

PPP Features

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

Fragment Size

1536 kbps	5 us	320 us	640 us	1.28 ms	2.56 ms	5.12 ms	10.24 ms
2024 kbps	4 us	256 us	512 us	1 ms	2 ms	4 ms	6 ms

The overall serialization latency for a fragment over a synchronous/ asynchronous Serial or T1 link should be multiplied by the size of the transmission queue. To control latency, both the transmission queue size and fragment size must be controlled.

Fragment Interleaving Over the Link

Transmitting a higher priority packet, either voice or video, with minimum and controlled latency, especially when lower priority packets such as data packets are present, is performed by the XSR automatically interleaving the fragment of the higher priority packet into the stream of the fragment of the lower priority packet. Fragment interleaving ensures that the transmission latency of a higher priority packet will not exceed the maximum value as determined by the sum of the fragment and transmission queue sizes.

To sum up, Multi-Class MLPPP performs as follows:

•Each packet is designated with a class number based on the classification result of the packet type. Packet classification is performed by the external method like QoS, which classifies the packet according to the ACL and assigns a class number according to priority as follows: the smaller the class number, the lower the priority. The current class number ranges from 1 to 4, with 1 for lower priority packets and 4 for the highest. Class 0 is reserved for a packet in the fair-queue class.

•Each packet is split into the small size. The maximum fragment size is defined as the maximum delay allowed over the slowest speed link in the bundle.

•Each fragment is appended with the MLPPP header, which includes the class information and sequence number dedicated to this class and transmitted over the member links. Fragments of different classes are interleaved.

•Fragments with a higher class number suspend the transmission of fragments with a lower class number that is pending for transmission, which is the interleaving of the fragments with different class numbers.

•When fragments are received, they are re-assembled based on class and sequence number with the order of the packet within the same class preserved.

Multilink Head Format Negotiation

The Multilink Head Format is negotiated between peers based on whether an MRRU, short sequence number, and code are present, as listed in Table 8-2on page 8-7:

Table 8-2 Multi-Class MLPPP Negotiation

		Option Type
MRRU	Short Sequence # Multilink Head Format		Negotiation Result
present	nil	nil	MLPPP with Long Sequence # & No Multi-Class
present	present		MLPPP with Short Sequence # & No Multi-Class
present	nil	present/Code = 2	MLPPP with Long Sequence # & Multi-Class

XSR User’s Guide 8-7