Critique paper on wireless network research paper

Paper Critique

Title: BBN: Throughput Scaling in Dense Enterprise WLANs

with Blind Beamforming and Nulling

Author:

Wenjie Zhou, Tarun Bansal, Prasun Sinha and Kannan Srinivasan

Publication: Proceeding of the 20th annual international conference on Mobile computing and networking (MobiCom), 2014.

Reviewer:

XXXXXX

Date:

XXXXXX

In this paper, the author realizes that traditionally WLAN design pays less attention to the uplink traffic

which in contrast increases rapidly due to the vast mobile applications recently. The paper aims to scale

the uplink throughput with the number of clients. The proposed BBN (blind beamforming and nulling)

leverages the high density of access points (APs) in Enterprise WLANs (EWLANs) to beamforming the

uplink traffic simultaneously in only 2 time slots. Then, the throughput is improved compared with

traditional TDMA and 802.11 schemes. The author also claims their design holds the merits such as

shifting more computation and design complexity to APs and exchanging less overhead in the backbone

than existing efforts. Practical issues are also considered and solved to some extent.

Strengths:

The topic is timely and a rise in attention of the uplink traffic in recent communication. The approach

leveraging the nature of EWLANs is novel. In mathematics, the proposed protocol is proven to be very

efficient in achieving scaled throughput when there are sufficient APs. The algorithm does require less

operation on mobile clients. The paper is overall well written with clear logic.

Weaknesses:

Some practical issues mentioned in the paper are actually crucial to this scheme, but the author failed to

provide the convincing solutions which put the practicability of BBN to question. Also, some underlying

assumptions are oversimplified, which cannot highlight BBN from the cited related works. However, the

performance results skip the comparison with these contenders. Moreover, these related works to justify

are not the state-of-the-art. Lastly, the name of the approach, BBN, is not so properly, since throughout

the paper the blind feature is questionable.

Detailed comments: First, the practical challenges in this paper are claimed to be solved in a tricky way. For example, 1) the

author introduces the techniques from other papers to achieve the synchronization of APs but without the

import of the overheads following these solutions; 2) to expand BBN to multi-collision domain, APs who

can hear each other are formed into groups and remain silent while the neighboring group is

communicating. This design is not novel and usually requires lots of backbone overheads; and 3) the

nature of beamforming leads to a strictly low endurance of decoding error (once a packet is decoded

incorrectly, all other beamforming packets are ruined). The extra AP solution cannot solve the problem

effectively, which can be seen from Fig. 8(c). However, the author did not take the correct decoded

packet into account in terms of the throughput, which is unfair when comparing it with other protocols

with low error decoding rate.

Then, some latent assumptions seem to indicate that the BBN is not blind. For instance, in order to

align the packet for off cutting, which packet is decoded at which AP and which subset of APs act as

transmitters are supposed to be known for every other AP. In other words, each AP has a good view of

both the clients and the role of other APs, which is hard for me to admit that this is called blind

beamforming. Even if the information is gained by estimating the expected highest SNR as claimed by

the author, this means each AP has an accurate pre-knowledge of the clients’ mobile topology, which is

an oversimplified assumption.

Further, the performance results are not sufficient to justify its conclusions: 1) the comparison is

incomplete. Author only compares its design with the general TDMA and 802.11 protocols. The previous

mentioned works are not considered in this part. Though they have the drawbacks as the paper criticized,

the BBN also has its own drawbacks that those papers do not have; 2) some metrics in this paper are

missing, such as the backbone overhead and error decoding rate as mentioned before. It is unfair to just

compare the throughput without these two metrics. Since the author claims it has less overhead in

backbone than the sample-sharing design, these two works should be compared on the overhead. Though

APs only exchange the decoded packet, I feel like BBN requires lots of other information exchanging

through the backbone while reading the paper; and 3) the performance of the multi-collision domain BBN

is not shown on the experiment. On the other hand, the assumption of the trace-driven simulation is not

clear.

At last, the related works ([18], [24], [7], and [9]) the paper used are too early works to justify BBN’s

salient features, since they are all papers before 2008 while BBN is a 2014 paper.

Conclusion: Paper attempts to provide throughput scaling linearly with the number of clients for EWLANs, which

is an ambitious goal. This would be great, if the protocol were not based on oversimplified assumptions

and the performance results over practical issues were given.