MACAW is an extension/modification to MACA, Media Access Collision Avoidance protocol earlier proposed by other researchers. The basis of these protocols is that the contention is at the receiver and not the transmitter pointing to the familiar "hidden" and "exposed" terminal problems and therefore extend on explicit contention management. Interestingly after 15 years from the publication date of this paper most of the WLAN technologies use carrier-sense and do not use such RTS/CTS mechanisms even though it is part of the standards. So it was interesting for me to understand what where their assumptions and arguments back then and why this hasn't been really accepted and deployed widely up to now.
MACAW was designed for a single channel WLAN operating at 5GHz with 256Kbps link radios. The structure of the target WLAN was a nano-cellular structure with small cell sizes (6m diameter) and single base station per cell. The radios are quite short range 3~4 meters. The authors claim many assumptions about the indoor environment under investigation which sound quite strange knowing what we know about indoor environments today:
The addition of ACK ensures the unreliable link doesn't wastefully reduce the TCP performance and this can be considered as one the most significant contributions of this paper. Another interesting result of this paper is that in order to have fair bandwidth allocation, contention estimations should be done collectively and individual approach will fail to reach this purpose. Furthermore a Multiplicative Increase Additive Decrease (MILD) scheme is proposed instead of the Binary Exponential Backoff (BEP) for adjusting the backoff parameters and shown to achieve better and fairer throughput when used along the collective contention estimation mechanisms.
The authors mainly compare their algorithm with MACA showing that the extra overhead is negligible compared to the gains they achieve and they mention that they achieve 78% of the link capacity. It is important to note that these calculation were done for 512 byte packets and if one considers smaller packets such as 802.15.4 128byte packets you will end up with almost 50% channel capacity! Never the less if the target system was designed only for large packets sizes this is a valid claim.
In addition the authors do not compare their scheme fairly with a CSMA/CA scheme. By discarding the carrier-sense method they fail to investigate the cases in which CSMA can achieve much better performance than explicit contention management. The overhead corresponding to MACAW like schemes is not negligible in general. Along this paper it looked like the authors need to introduce more and more control packets which adds to the overhead and complexity of this scheme. I believe the complexity in addition to the large overhead for small packets has been the main road blocks for large deployments of such schemes in WLANs today.
MACAW was designed for a single channel WLAN operating at 5GHz with 256Kbps link radios. The structure of the target WLAN was a nano-cellular structure with small cell sizes (6m diameter) and single base station per cell. The radios are quite short range 3~4 meters. The authors claim many assumptions about the indoor environment under investigation which sound quite strange knowing what we know about indoor environments today:
- Attenuation was considered a strong function of distance.
- The communication links were considered symmetric.
- The communication range was modeled as a binary in-range/out-of-range model and the other effects was ignored.
The addition of ACK ensures the unreliable link doesn't wastefully reduce the TCP performance and this can be considered as one the most significant contributions of this paper. Another interesting result of this paper is that in order to have fair bandwidth allocation, contention estimations should be done collectively and individual approach will fail to reach this purpose. Furthermore a Multiplicative Increase Additive Decrease (MILD) scheme is proposed instead of the Binary Exponential Backoff (BEP) for adjusting the backoff parameters and shown to achieve better and fairer throughput when used along the collective contention estimation mechanisms.
The authors mainly compare their algorithm with MACA showing that the extra overhead is negligible compared to the gains they achieve and they mention that they achieve 78% of the link capacity. It is important to note that these calculation were done for 512 byte packets and if one considers smaller packets such as 802.15.4 128byte packets you will end up with almost 50% channel capacity! Never the less if the target system was designed only for large packets sizes this is a valid claim.
In addition the authors do not compare their scheme fairly with a CSMA/CA scheme. By discarding the carrier-sense method they fail to investigate the cases in which CSMA can achieve much better performance than explicit contention management. The overhead corresponding to MACAW like schemes is not negligible in general. Along this paper it looked like the authors need to introduce more and more control packets which adds to the overhead and complexity of this scheme. I believe the complexity in addition to the large overhead for small packets has been the main road blocks for large deployments of such schemes in WLANs today.
No comments:
Post a Comment