Browse Prior Art Database

Link adaptation and RLC retransmission schemes for real-time traffic over GPRS/EDGE

IP.com Disclosure Number: IPCOM000129195D
Original Publication Date: 2005-Sep-30
Included in the Prior Art Database: 2005-Sep-30
Document File: 3 page(s) / 71K

Publishing Venue

Motorola

Related People

Suresh Kalyanasundaram: AUTHOR [+2]

Abstract

Real-time traffic such as streaming audio and video, push-to-talk, etc. are characterized by tight delay constraints, but somewhat lax packet loss constraints. But they are constrained into using one of the two existing RLC (radio link control) protocol modes, namely acknowledged or unacknowledged mode. In the acknowledged mode of operation, the RLC protocol does not move forward unless it receives acknowledgement for each and every RLC block. In the unacknowledged mode of operation, the RLC protocol transmits each RLC block just once and does not wait for an acknowledgement.

This text was extracted from a Microsoft Word document.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 42% of the total text.

Link adaptation and RLC retransmission schemes for real-time traffic over GPRS/EDGE

Suresh Kalyanasundaram

Rajeev Agrawal

1          Introduction

Real-time traffic such as streaming audio and video, push-to-talk, etc. are characterized by tight delay constraints, but somewhat lax packet loss constraints. But they are constrained into using one of the two existing RLC (radio link control) protocol modes, namely acknowledged or unacknowledged mode. In the acknowledged mode of operation, the RLC protocol does not move forward unless it receives acknowledgement for each and every RLC block. In the unacknowledged mode of operation, the RLC protocol transmits each RLC block just once and does not wait for an acknowledgement.

Because of the tight delay constraints, we will not be able to operate RLC in acknowledged mode. Therefore, we will be forced into using RLC in unacknowledged mode. From the point of view of minimizing resource usage (and hence maximizing capacity), it is known that, in adaptive modulation and coding systems, the jump to a higher modulation and coding scheme happens when the block error rate is fairly large. If the link adaptation algorithm that minimizes resource consumption for acknowledged mode is used, the observed block error rate can be as large as 20% in the worst case. Moreover, since the entire packet is lost even if one of the blocks is in error, the effect gets magnified even more from the application layer. For example, our simulations indicate that the packet error rate can be as high as 17% for push-to-talk traffic operating with the current resource minimizing link adaptation algorithm for acknowledged mode operation.  Such a high packet error rate will result in very poor quality application performance.

The reason for this problem is that the link adaptation algorithm has been designed for acknowledged mode and not for unacknowledged mode. Our invention is the design of a link adaptation algorithm that minimizes resource consumption (and hence maximizes capacity) of the system for real-time traffic operating in unacknowledged mode as well.

From the point of view of maximizing the capacity of the system for real-time traffic, we should operate RLC in a different mode where each RLC packet is retransmitted a fixed number of times and no more. The RLC protocol should transmit a given RLC block either until it receives an ack or for the maximum number of times allowed. The protocol should stop retransmitting that block and move on to subsequent blocks if it turns out that the maximum limit on the number of retransmissions is reached. The number of retransmissions that we are willing to tolerate will be determined by the end-to-end application tolerance for delay.

Our link adaptation algorithm has been designed for this above mode of RLC, and a special case of which is the unacknowledged mode of operation where the number of allowed retransmissions is zero. The main idea behind our link adaptation algorithm is that we minim...