QoS based scheduling in the downlink of multi-user wireless systems (extended)☆
Introduction
IEEE 802.11n [1] is proposed as an amendment of the previous IEEE 802.11 wireless networking standard to significantly improve network throughput. It aims at providing a data transmission rate of up to 600 Mbps. The version 2.0 draft specification for the next generation IEEE 802.11n WLANs has been approved in March 2007. It has introduced substantial enhancements at both the PHY (physical) and MAC (medium access control) layers for high throughput, efficiency and robustness [2], [3] for the wireless system. In the PHY layer, based on the MIMO-OFDM (Multiple Input Multiple Output – Orthogonal Frequency Division Multiplexing) technology, 802.11n can use spatial multiplexing to transmit two or more data streams simultaneously. It also provides transmitter spatial diversity to improve reception by spreading the spatial streams across multiple antennas [4]. Beamforming, specified as an optional feature, can further improve packet transmission efficiency. The 802.11n defines a new set of the modulation and coding schemes (MCS), and the MCS is an index value that determines the modulation, coding and number of spatial streams in MIMO-OFDM systems. The actual transmission scheme is composed of both the MIMO mode and the MCS. The efficiency improvements at the MAC layer are frame aggregation, block acknowledgment (block ACK, also backward compatible with 802.11e [5]), etc. Frame aggregation [6], [7], [8] can improve MAC-layer efficiency by reducing the transmission time for preamble and frame headers, and the random backoff period for successive frame transmissions. They are particularly applicable to voice traffic where the voice frame is short and continuous traffic such as video or large file transfers.
We tackle the wireless scheduling problem from a cross layer optimization angle. We have done much research on the link adaptation algorithms for opportunistic scheduling in [9]. However, while the link adaptation improves transmission on a physical link, the aggregate system performance is very much dependent on multi-user scheduling and cross layer optimization mechanisms, which are also heavily coupled with underlying link adaptation. This cross layer optimization becomes more imperative in the 802.11n wireless systems since the standard has introduced many significant options in the MAC layer. In this paper, we will take advantage of the frame aggregation in the 802.11n for designing a multi-user scheduling algorithm. The scheduler tries to improve the system performance, in particular, in terms of the quality of service (QoS) efficiency.
Different from the well studied opportunistic scheduling that monitors the channel continuously and decides the locally optimized strategy to send packets to one user, scheduler with frame aggregation can send packets to several users simultaneously. How to select this set of users to be serviced is a challenging problem, which can be easily modeled into a knapsack problem that is well known to be NP hard [10]. We argue that, however, in practice we do not have to find the optimal solution to the knapsack problem to achieve a good performance. A simple greedy algorithm that has less complexity works sufficiently well with practical implementation of the frame aggregation.
The rest of this paper is organized as follows. In Section 2, we present the system model and conventions used throughout the paper. We then model the scheduling problem in the wireless system with frame aggregations into a knapsack problem in Section 3. In Section 4, we propose a simple greedy algorithm to do the scheduling and analyze its performance. In Section 5, we bring forward some practical considerations and argue that the greedy algorithm performs well in practice. In Section 6, some simulations and performance evaluation are present comparing our algorithm with the Round-Robin scheduling. Then, we conclude the paper in Section 7.
Section snippets
System model and conventions
In this section, we first describe the system model under consideration and put forward the scheduling problem. We also model the general frame aggregation scheme and formulate users’ simple QoS requirements.
Knapsack problem and the largest unit urgency first scheduling
In each frame cycle, the scheduler is responsible to select a set of users to grant receiving packets.
With the modeling of users’ urgency, it is obvious for a good scheduler to maximize the total users’ urgency in each frame. Since the frame size is bounded by , we formulate the scheduler problem as follows:where , for . If , user i is selected for current frame cycle; otherwise it is not.
This is a combinatorial
Analysis of the largest unit urgency first scheduler
We first analyze the complexity of LUUF. It is easy to see that the complexity is since the LUUF basically involves two phases: sorting the unit urgencies and selecting users. It is well known that the sorting complexity is [10] and the selecting process costs since it is a sequential process. Overall, the LUUF complexity is .
To analyze the performance of LUUF, we should find out how close the solution given by LUUF is near the optimal solution.
We first define an
Discussions
In the previous section, we analyzed the LUUF performance under the assumption that the frame size is fixed at . In fact, in practice we can have a more flexible scheduler. According to the LUUF, if the selected users only occupy a frame size of , it is unnecessary for LUUF to wait for a period of to start next frame cycle. Therefore, we can adjust current frame size to and transmit packets according to the results of LUUF. According to argument 1 in Section 4, we know
Simulations and performance evaluation
In this section, we conduct some simulations comparing our scheduling algorithm LUUF with the Round-Robin scheduling algorithm. Round-Robin is a simple but widely used scheduling algorithm. In order to decide which users will get served in the next time slot, the Round-Robin scheduler keeps all users in a circular queue and visits them one by one. If adding the current user’s frame does not lead to overflow the maximum aggregated frame size , the current user will be served. Otherwise, the
Conclusions and future work
In this paper, we investigate the multi-user scheduling in the MAC layer of wireless systems with support of frame aggregation. We first model the scheduling problem into a knapsack problem that is NP hard and computational intractable. We propose a simple and efficient algorithm (LUUF) to approximate the optimal solution. Our analysis shows that LUUF exhibits rather good performance when combined with practical considerations. In particular, the LUUF works better with variable frame size and
References (12)
IEEE 802.11n: enhancements for higher throughput in wireless LANs
IEEE Wireless Commun.
(2005)- IEEE 802.11 WG, Draft PAR for High Throughput Study Group, May...
- IEEE 802.11n/D2.0, Draft amendment to STANDARD [for] information technology-telecommunications and information exchange...
- et al.
A high-performance MIMO OFDM wireless LAN
IEEE Commun. Mag.
(2005) - L.A.N. Wireless, Medium Access Control (MAC) and Physical Layer (PHY) specifications, part 11, IEEE Std., vol. 802.11e,...
- et al.
IEEE 802.11n MAC frame aggregation mechanisms for next-generation high-throughput WLANS
IEEE Wireless Commun.
(2008)
Cited by (4)
Dynamic multiuser sub-channels allocation and real-time aggregation model for ieee 802.11 wlans
2014, IEEE Transactions on Wireless CommunicationsPerformance analysis of fair scheduler for A-MSDU aggregation in IEEE802.11n wireless networks
2014, 2014 2nd International Conference on Electrical, Electronics and System Engineering, ICEESE 2014A-MSDU real time traffic scheduler for IEEE802. 11n WLANs
2012, IEEE Symposium on Wireless Technology and Applications, ISWTAFrame aggregation in wireless networks: Techniques and issues
2011, IETE Technical Review (Institution of Electronics and Telecommunication Engineers, India)
- ☆
This research work is partially funded by The Hong Kong Research Grants Council under the Grant RGC 610307.