Fragmentation/Aggregation Scheme for Throughput Enhancement of IEEE 802.11n WLAN Alexey Sidelnikov§ , Jeonggyun Yu† and Sunghyun Choi‡ § Samsung Electronics, Co., LTD †‡ School of Electrical & Computer Engineering and INMC, Seoul National University Email: §
[email protected] †
[email protected] and ‡
[email protected]
Abstract— In this paper, we propose a simple fragmentation/aggregation scheme which combines the MSDU fragmentation (used to enhance the reliability at the cost of more protocol overhead) of the legacy 802.11 and A-MPDU aggregation (used to reduce the overhead) of the emerging high-speed IEEE 802.11n. The proposed scheme increases the transmission range (i.e., dramatically increased throughput in bad channel condition) of high data rates via fragmentation without severely increasing the protocol overhead thanks to aggregation.
I. I NTRODUCTION Recently, along with many emerging applications and services over IEEE 802.11 WLAN, the demands for the faster and higher capacity WLANs have been increasing drastically. However, the 802.11 is known to have a high overhead for the MAC/PHY operations. It was shown in [6] that by simply increasing the PHY rate without reducing the medium access control (MAC)/physical (PHY) layer overhead, the enhanced throughput is bounded under 100 Mb/s, even if the PHY rate goes to infinity. It means that we need to enhance the 802.11 MAC by reducing overheads in order to create the next-generation high speed WLAN. In [7], Kim et al. showed the throughput performance of WLAN can be significantly enhanced via the frame aggregation. However, generally, the frame error rate (FER) increases as the size of the frame increases. As we can see in [7], the increasing FER can shrink the transmission range. Specifically, the more frames are aggregated, the smaller the transmission range is. In this paper, we propose a simple fragmentation/aggregation scheme, which combines the MAC Service Data Unit (MSDU) fragmentation of the legacy 802.11 and Aggregate-MAC Protocol Data Unit (A-MPDU) aggregation scheme of the emerging high-speed IEEE 802.11n MAC protocol [4]. II. P ROPOSED F RAGMENATION /AGGREGATION S CHEME IEEE 802.11 MSDU fragmentation leads to a very severe performance degradation especially at high data rates [5]. The reason is that though the fragment transmission time decreases, the PHY layer overhead is fixed, and hence, it becomes relatively larger and larger as data rate and/or the number of fragments increases. Therefore reducing the overhead is important to enhance throughput performance. On the other hand, fragmentation increases the reliability by increasing the probability of successful transmission of a number of short MPDUs in cases where channel characteristics limit the reliable reception for longer frames. As a result, there is always a tradeoff between reliability (i.e., fragmentation) and throughput for WLAN.
Fig. 1.
MAC header compression
The proposed fragmentation/aggregation scheme combines the very efficient bandwidth utilization of the frame aggregation scheme with the robustness to the channel errors from fragmentation process. As we can see in Fig. 2, the main idea is to fragment an MSDU from the logical link control (LLC) into several MPDUs and then transmit all of them in an A-MPDU container, i.e., as one PHY frame. 802.11e Block Acknowledgement (BA) mechanism is used for acknowledgements. Together with 802.11e BA mechanism being used for acknowledgements, this significantly reduces the PHY layer overhead compared with the distributed coordination function (DCF) fragmentation, and increase reliability compared with no-fragmentation case. In the case when the transmission opportunity (TXOP) limit is large enough, more than one MSDU can be accommodated in one A-MPDU. Fig. 2 illustrates that each MSDU is divided into two equal-sized fragments, i.e., MPDU 1∼2 and MPDU 3∼4 are from different MSDUs. A cross represents an MPDU transmission failure due to the channel errors. When a fragment fails, but the block acknowledgement request (BAR)/BA frame exchange is successful, then according to the EDCA backoff rule [3], the backoff window size is reset to the minimum value. However, if the BAR/BA exchange fails, the sender should increase the backoff window size, and retransmit a BAR frame until it receives a BA successfully. After that, if there is still a residual TXOP duration, the sender will transmit more MPDUs in A-MPDU. For the given example, it does not matter whether an MPDU contains just a fragment of an MSDU or the whole MSDU. As we can see in Fig. 1, the MAC header contains a number of fields which are not required in every fragment, namely, Duration/ID, Addresses 1∼4 and QoS control fields. By removing them from all the fragments but the first one, a substantial throughput gain is achieved. The possible drawback is that if the first fragment from an MSDU as well as BAR control frame in an A-MPDU are corrupted, then the AMPDU frame will be dropped by the receiver. However, this probability is very negligible particularly due to the very small BAR frame size. We analyze several schemes including our scheme mathematically. However, we do not show the analysis in detail due to the page limit.
Fig. 2.
Timing diagram of the proposed fragmenation/aggregation scheme
III. P ERFORMANCE E VALUATION We, in this section, evaluate the performance of our proposed scheme. We plot analytical results with lines and simulation results with points. We validate our analytical model using ns-2 simulator [8], and the simulation results match the mathematical results very well. We assume a very simple network topology with a single transmitter with infinite traffic amount and a single receiver. We also assume the 802.11a PHY providing 6 to 54 Mb/s since the 802.11n PHY is not finalized yet. We use 1582 bytes long MSDUs, so each can be fragmented into seven 256 bytes long MPDUs, which is the minimum fragment size [1]. The TXOP duration is set to the 802.11e default value, i.e., 3.008 ms [3]. Fig. 3 shows the effective throughput performance vs. SNR for a fixed data rate of 54Mb/s in 802.11a PHY. The proposed scheme outperforms aggregation-only scheme for low SNR values. However, when the channel condition is very good, the proposed scheme delivers roughly 10% less throughput. But the use of MAC header compression reduces the fragmentation overhead almost to zero in that region, and provides around 0.6 dB SNR gain in low SNR region. Fig. 4 presents the ideal (i.e., the best PHY mode available) link adaptation [5] for all the schemes described above. The results for 802.11 DCF and 802.11e EDCA schemes are also presented for the comparison purpose. As can be observed, the proposed scheme works the best in the transition region, i.e., where frame error rate is not close to zero. Then, the achievable SNR gain is as high as up to 0.6 dB. At the same time, the use of MAC header compression significantly reduces the performance loss due to fragmentation. IV. C ONCLUSION In this paper, we proposed a simple fragmentation/aggregation scheme, which combines the MSDU fragmentation of legacy 802.11 and aggregation scheme, i.e., A-MPDU scheme from the emerging high-speed IEEE 802.11n MAC protocol. Moreover, we showed via both analysis and simulation results that the proposed scheme basically increases the transmission ranges for high data rates via fragmentation. At the same time, the overall performance of the proposed scheme does not suffer from fragmentation overheads. R EFERENCES [1] IEEE Std. 802.11, Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) specifications, Aug. 1999.
Fig. 3. Throughput performance of the proposed scheme; 54 Mb/s data rate.
Fig. 4. Throughput performance of the proposed scheme; various data rates. [2] IEEE 802.11a, Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) specifications: High-speed Physical Layer in the 5 GHz Band, Sep. 1999. [3] IEEE 802.11e, Part 11: Wireless LAN Medium Access Control (MAC) and Physical layer (PHY) specifications: Medium Access Control (MAC) Enhancements for Quality of Service (QoS), Nov. 2005. [4] IEEE P802.11n/D1.0, Part 11: Wireless LAN Medium Access Control (MAC) and Physical layer (PHY) specifications: Enhancements for Higher Throughput, Mar. 2006. [5] D. Qiao and S. Choi, “Goodput Enhancement of 802.11a Wireless LAN via Link Adaptation,” Proc. IEEE ICC’01, Helsinki, Finland, June 2001. [6] Y. Xiao, and J. Rosdahl, “Throughput and Delay Limits of IEEE 802.11,” IEEE Communications Letters, vol. 6, no. 8, Aug. 2002. [7] Y. Kim, S. Choi, K. Jang, and H. Hwang, “Throughput Enhancement of IEEE 802.11 WLAN via Frame Aggregation,” in Proc. IEEE VTC’04Fall, Los Angeles, Sept. 2004. [8] “The Network Simulator – ns-2,” http://www.isi.edu/nsnam/ns/, online link.
