Determine the Candidates for Passive Scan to Get More Benefit
To simplify the derivation of possible solutions, some channels are selected as the candidates for passive scan. That is, the channels which few APs are configured on are selected first, thus only few beacons need to be heard. It is said that a channel has a short channel occupied time when STAs only need to wait on that channel for a few time for scanning. Therefore, the shorter the channel occupied time is, the fewer influence it has on the real-time services. The channels are discarded from the candidate list when they have channel occupied time longer than the time when performing a normal active scan. An example is illustrated in Fig. 8. A channel with too many APs may have a channel occupied time longer than a single-channel active scan time and will be discarded from the candidate list. The occupied time test reduces the number of potential passive-scan channels which have more possibility to satisfy the delay requirements.
Fig. 8. Occupied time test
The selected candidates are then sorted by the number of APs. Thus the channels with fewer channel occupied time are scheduled first to reduce the influence on real-time services and also the potential collisions with other potential channels.
Adjusting After Scheduling
After deciding the scanning type of each channel, an adjusting function is done to further improve the scan performance. The adjusting function checks the scheduled list and tries to replace the passive-scan slots of a channel with a single active scan slot to shorten the total turnaround time. Fig. 9 presents the improvement of replacing the passive-scan slots of channel 4 with a single active scan slot. The adjusting function checks the time slots in the scheduled list from tail to head until the time slot checked cannot be replaced by any other earlier time slot. Through this mechanism, the resource utilization is improved during the
scanning time while the delay constraints are satisfied. The total turnaround time is shortened as well.
Fig. 9. Adjustment
The Algorithm
The function INSERT_TO_SL mentioned above is used to insert a passive-scan channel into the scheduled list. The detail algorithm is described as follows.
BUILD_CANDIDATE_LIST(ENV)
Input. ENV: environment information which contains channel elements Output. CandiList: candidate list of potential passive-scan channels for each CHi in ENV
1
do
2
Calculate Tact, which is the minimal turnaround time to finish a standard active scan on CHi
3
under delay constraints;
4
Calculate the channel occupied time for CHi;
5
if the channel occupied time of CHi is shorter then Tact
6
then
7
Insert CHi to CandiList and sort the list by the number of APs;
8
end
9
Return CandiList;
10
ASA_FAF(ENV)
Input. ENV: environment information which contains channel elements Output. SL: the result scheduled list
ScanTime: the total turn-around time for scanning
Call BUILD_CANDIDATE_LIST to list the potential passive-scan channels which are sorted
1
according to the number of APs;
2
for each CHi in CandiList
3
do
4
Call INSERT_TO_SL to insert CHi to the scheduled list;
5
Remove redundant beacon slots to keep each AP scanned only once;
10
Insert the active slot to SL for each channel marked as SCAN_ACT;
11
Call ADJUST to further improve the final result;
12
Calculate ScanTime as the total turnaround time according to SL;
13
Return SL and ScanTime;
14
First a candidate channel list is built. Channels in this list have shorter channel-occupied time when performing passive scan rather than active scan. The less time-occupied channels are scheduled first, thus the candidates are sorted according to their number of APs, which is from fewest to most. Then the candidates are checked one by one, until all of them are scheduled. The channels failed to join the scheduled list with passive scan are marked as active scan. Since there may be more than one beacon slots from a single AP are inserted into the scheduled list, the algorithm checks the time slots and removes the redundancies. Then active-scan slots are inserted for the other un-scheduled channels. After all channels are scheduled, the adjusting function is called. The scheduled result is modified to get better
resource utilization and time performance.
This heuristic solution has a time complexity of O(Nbeacons2
), which is much lower than the time complexity of the optimal solution.
V. SIMULATION AND EVALUATION
Simulation Settings
In this section, six scanning mechanisms, including two standard scanning mechanisms, two modified mechanisms as mentioned in the related work section, and the proposed ASA algorithms are analyzed. The modified mechanisms are named preactive active scan (preACT) and preactive passive scan (prePAS), which take advantage of the environment information to reduce the channel-waiting time. In the preACT mechanism, the NIC skips the empty channels rather than waits until minimum channel time. In the prePAS mechanism, the NIC knows the beacon arrival time of each AP and hears the beacons exactly whenever they arrive.
Both of these two modified mechanisms satisfy the user-defined delay constraint. The optimal and heuristic ASA algorithms are implemented and can be used directly in the driver to arrange the scanning phase.
A virtual background environment is generated. The number of total APs from one to ten are randomly distributed over 11 channels. Each AP has the first beacon arriving randomly within 100 ms and periodically arrives again every beacon interval, which is set as 100 ms.
Voice packets are sent every 20 ms and user-defined maximum delay constrains are set to 20 ms, 60 ms, 120 ms, and infinite for each case. An average of 1000 runs is taken for the performance evaluation.
Other parameters for simulation are as follows. Channel switching time is set as 5 ms
according to the measurement in [10]. Min/max-channel time for active scan are 1 and 11 ms.
Time to receive a beacon is no more than 1 ms according to [16]. The short-time scale measurement indicates that the clock drift in APs is negligible [10].
Performance Evaluation
The six mechanisms are performed under the delay constraint of 20 ms to analyze the total scan time, as shown in Fig. 10. When nAp=10, the optimal solution OPT has the best performance of 93.5 milliseconds in average. The FAF mechanism shows a result of 100.4 ms which is close to the OPT solution. The standard passive scan waits for a beacon interval on each channel, thus has a total scan time of 1055 ms, no matter how many APs are in the background. When there is only one AP in the background, which is also the associated AP, all mechanisms except for the standard ones don’t have to spend any time on scanning because of the awareness of background information, thus the total scan time is zero in these cases. Besides, since the performance of prePAS mechanism depends on the beacon arrival time, it has a total scan time as long as 213.3 ms when there are 10 APs in the background.
Fig. 10. Total scan time for each mechanism
The enhancement comparing to the standard active scan mechanism is shown in Fig. 11.
The FAF mechanism has an enhancement of 30.06% when nAp=10, which is much better than the preACT and prePAS mechanisms and very close to the enhancement of the OPT solution, which is 34.87% better than the standard active scan.
Fig. 11. Enhancement comparing to active scan under maxDelay=20ms
Fig. 12 shows the cumulative distribution of packet delay of each mechanism. The maximum delay is set to 20 ms. More than 90% of voice packets have negligible delay (<1ms) in both OPT and FAF mechanisms, which obviously outperform the standard mechanisms and the preACT solution.
Fig. 12. Delay distribution of each mechanism under maxDelay=20ms
In the last part of this section, the execution time of FAF and OPT scheduling is compared. The two algorithms are run on a personal computer with AMD Athlon 1.83GHz cpu. The result is as shown in Fig. 13. The execution time of the optimal solution increases exponentially through the number of total APs increases. When there are 10 APs in the background, the optimal solution has a schedule latency of 59.84 ms which is much longer than the FAF algorithm.
Fig. 13. Execution time for scheduling
VI. CONCLUTION
By combining active and passive scan strategies, the proposed ASA mechanism minimizes the total scan time without introducing packet loss, delay and jitters. Simulation results show that about 30% scan time can be reduced while the QoS requirement of the connection can be guaranteed during handoffs. A heuristic algorithm was further presented to reduce the computation complexity of the proposed ASA approach. The simulation results also demonstrate that the heuristic algorithm can be easily implemented and achieves near-optimal solutions.
References
[1] Sean Christensen, “Voice over IP solutions,” Juniper Networks, June, 2001.
[2] ETSI, European Telecommunications Standards Institute, http://www.etsi.org/
[3] Alan Sicher and Randall Heaton, “GPRS technology overview,” Dell, February 2002.
[4] IEEE, “Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications,” IEEE standard 802.11, 1999.
[5] IEEE, “Trial-Use Recommended Practice for Multi-Vendor Access Point Interoperability via an Inter-Access Point Protocol Across Distribution Systems
Supporting IEEE 802.11 Operation,” IEEE standard 802.11f, July 2003.
[6] IEEE, “Draft Amendment to Standard for Information Technology - Telecommunications and Information Exchange Between Systems - LAN/MAN Specific
Requirements - Part 11: Wireless Medium Access Control (MAC) and Physical Layer
(PHY) specifications: Amendment 2: Fast BSS Transition,” P802.11r D1.0, November
2005.
[7] IEEE, “Media Access Control (MAC) Enhancements for Quality of Service (QoS),”
IEEE Standard 802.11e/D6.0, December 2003.
[8] IEEE, ”Radio Resource Measurement Enhancements,” IEEE 802.11k D1.0, September 2004.
[9] A. Mishra, M. Shin, and W. A. Arbaugh, “An Empirical Analysis of the IEEE 802.11
MAC Layer Handoff Process,” ACM SIGCOMM Comput. Commun. Rev. 33, 2, pp.93 –
102, April 2003.
[10] Ishwar Ramani, and Stefan Savage, “SyncScan: Practical Fast handoff for 802.11 Infrastructure Networks,” in Proc. IEEE INFOCOM 2005, March 2005.
[11] Matthew S. Gast, “802.11Wireless Networks- The Definitive Guide,” O’REILLY, 2005.
[12] Marc Emmelmann. "Influence of Velocity on the Handover Delay associated with a Radio-Signal-Measurement-based Handover Decision," In Proc. of IEEE Vehicular
Technology Conference (VTC 2005 Fall), September 2005.
[13] Minho Shin, Arunesh Mishra, and William A. Arbaugh, “Improving the Latency of 802.11 Hand-offs using Neighbor Graphs,” ACM MOBISYS 2004, June 2004.
[14] S. Pack and Y. Choi, “Fast Inter-AP Handoff Using Predictive Authentication Scheme in a Public Wireless LAN,” in Proceedings of IEEE Networks Conference, Aug.
2002.
[15] A. Markopoulou, F. A. Tobagi, and M. J. Karam, “Assessment of VoIP quality over internet backbones,” IEEE INFOCOM 2002 - the Conference on Computer
Communications, June 2002.
[16] A. Trad, F. Munir, and H. Afifi, "Capacity Evaluation of VoIP in IEEE 802.11e Network Environment", CCNC'06 - IEEE Consumer Communications and Networking
Conference, January 2006.