ARROW: An Efficient Traffic Scheduling Algorithm for ... - IEEE Xplore

3 downloads 0 Views 567KB Size Report
Abstract—In this paper we present a novel traffic scheduling algorithm for IEEE 802.11e, referred to as ARROW(Adaptive. Resource Reservation Over WLANs), ...
3558

IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 5, NO. 12, DECEMBER 2006

ARROW: An Efficient Traffic Scheduling Algorithm for IEEE 802.11e HCCA Dimitris Skyrianoglou, Student Member, IEEE, Nikos Passas, Member, IEEE, and Apostolis K. Salkintzis, Senior Member, IEEE

Abstract— In this paper we present a novel traffic scheduling algorithm for IEEE 802.11e, referred to as ARROW(Adaptive Resource Reservation Over WLANs), which aims at providing improved performance for the support of multimedia traffic. The novel characteristic of this algorithm, compared to previous proposals, is that it performs channel allocations based on the actual traffic buffered in the various mobile stations, i.e., on the exact transmission requirements. This feature renders ARROW ideal for variable bit rate traffic. However, an enhancement is also presented that improves ARROW performance under constant bit rate traffic. The ARROW algorithm and its enhancement are discussed and evaluated against two other schedulers found in the open technical literature, namely the Simple Scheduler and SETT-EDD. Results from a detailed simulation model verify that ARROW provides much better channel utilization and considerably improved performance, in terms of mean delay and packet loss. Index Terms— Multimedia QoS, 802.11e, traffic scheduling.

I. I NTRODUCTION HE IEEE 802.11 standard [1] is considered today the dominant technology for wireless local area networks (WLANs). Besides great research interest, 802.11 has enjoyed widespread market adoption in the last few years, mainly due to low-price equipment combined with high bandwidth availability. Recent improvements in the physical (PHY) layer provide transmission speeds up to hundreds of Mbps per cell, facilitating the use of broadband applications. However, one of the main weaknesses of the original 802.11, towards efficient support of multimedia traffic, is the lack of enhanced Quality of Service (QoS) provision in the Medium Access Control (MAC) layer. The mandatory access mode in 802.11, referred to as Distributed Coordination Function (DCF), provides contention-based access to the wireless medium, while the optional Point Coordination Function (PCF) allows contentionfree access, but has no means to provide tight control on QoS metrics. In order to eliminate these weaknesses and respond to business requirements for multimedia over WLANs, IEEE

T

Manuscript received October 8, 2004; revised July 13, 2005 and October 25, 2005; accepted November 17, 2005. The associate editor coordinating the review of this paper and approving it for publication was K. K. Leung. This work was performed in the context of the project entitled “PYTHAGORAS: Support of Universities’ Research Groups” co-funded by the Operational Programme for Education and Initial Vocational Training (O.P. Education) and the European Social Fund. D. Skyrianoglou and N. Passas are with the Department of Informatics and Telecommunications, University of Athens, 32 Kifissias Ave., Athens, Greece, 15784 (e-mail: [email protected]; [email protected]). A. K. Salkintzis is with Motorola, 32 Kifissias Ave., Athens, Greece, 15125 (e-mail: [email protected]). Digital Object Identifier 10.1109/TWC.2006.04669.

is currently working on a set of QoS-oriented specification amendments, referred to as IEEE 802.11e, that enhance the existing MAC protocol and facilitate the multimedia QoS provision. In IEEE 802.11e [2], the QoS mechanism is controlled by the Hybrid Coordinator (HC), an entity that implements the so-called Hybrid Coordination Function (HCF). The HC is typically located in an Access Point (AP) and utilizes a combination of a contention-based scheme, referred to as Enhanced Distributed Coordination Access (EDCA), and a polling-based scheme, referred to as HCF Controlled Channel Access (HCCA), to provide QoS-enhanced access to the wireless medium. EDCA provides differentiated QoS services by introducing classification and prioritization among the different kinds of traffic, while HCCA provides parameterized QoS services to Stations (STAs) based on their traffic specifications and QoS requirements. To perform this operation, the HC has to incorporate a scheduling algorithm that decides on how the available radio resources are allocated to the polled STAs. This algorithm, usually referred to as the “traffic scheduler”, is one of the main research areas in 802.11e, as its operation can significantly affect the overall system performance. In this paper, we propose and study a new scheduling algorithm, aiming at improving the performance over the existing schedulers. The novel characteristic of this algorithm, referred to as ARROW (Adaptive Resource Reservation Over WLANs), is that it allocates the available bandwidth based on the actual amount of data awaiting transmission in every STA. This is in contrast to previous proposals, which perform channel allocations based on the estimated buffered data in every STA. As we discuss below, every STA communicates the amount of its buffered data to the HC, by means of protocol elements already defined in the IEEE 802.11e standard. Therefore, ARROW does not mandate any protocol changes. The rest of the paper is organized as follows. Section II provides a brief description of the new access modes introduced by 802.11e, together with well-known traffic schedulers found in the technical literature. Section III contains a detailed description of ARROW, focusing on its advantages against existing proposals. Section IV describes the simulation model used for evaluating ARROW and presents extensive simulation results that validate the claimed performance improvements. Finally, Section V contains conclusions and plans for future work.

c 2006 IEEE 1536-1276/06$20.00 

SKYRIANOGLOU et al.: ARROW: AN EFFICIENT TRAFFIC SCHEDULING ALGORITHM FOR IEEE 802.11E HCCA

II. ACCESS M ODES IN IEEE 802.11 E As already mentioned, IEEE 802.11e introduces two new access modes, namely the Enhanced Distributed Coordination Access (EDCA) and the HCF Controlled Channel Access (HCCA), which are briefly described below. STAs and APs that support these access modes are referred to as QoS STAs (QSTAs) and QoS APs (QAPs) respectively. A QAP uses special purpose beacon frames to divide time into fixed length Beacon Intervals containing Contention Periods (CPs) and optional Contention-Free Periods (CFPs). EDCA is used in CPs only, while HCCA can be used in both periods. A basic concept in 802.11e is the Transmission Opportunity (TXOP), defined as an interval of time when a QSTA obtains permission to transmit onto the shared wireless channel. A TXOP is defined by a starting time and a maximum duration. TXOPs are obtained via contention in EDCA mode or are granted by HC in HCCA mode. The differentiated QoS offered in EDCA is realized through the introduction of eight Traffic Categories (TCs), each one characterized by a different User Priority (UP). Each UP is mapped to one out of four different Access Categories (ACs), each one having a particular set of access parameters that ultimately define the medium access priority. In the CP, the ACs within a QSTA independently start a backoff after detecting the channel being idle for a certain time interval. Higher priority ACs can start their backoff earlier than others to gain a time advantage. Each backoff sets a counter to a random number drawn from the range of its current Contention Window (CW). As in the legacy DCF, when the medium is determined busy before the counter reaches zero, the backoff has to wait for the medium to become idle again, before continuing to count down. When multiple ACs have buffered data for transmission, contention for the medium occurs both internally (within a QSTA) and externally (between different QSTAs), based on the same coordination function, so that internal scheduling resembles external scheduling. Internal collisions are resolved by allowing the higher priority queue to transmit, while the lower priority invokes a queue-specific backoff as if a collision had occurred. Several proposals can be found in the literature describing mechanisms that exploit the capabilities of EDCA, in order to differentiate service provided to different kinds of traffic (e.g., [3], [4], [5], [6]). Although proper use of EDCA can lead to improved performance, especially for high priority traffic, its contention-based operation cannot provide strict QoS control at all times. HCCA on the other hand, provides much more control of the wireless medium, by assigning contention-free TXOPs during both the CP and the CFP. TXOPs are assigned per QSTA, although the decision is based on the characteristics of individual Traffic Streams (TSs). A TS is a set of MAC Service Data Units (MSDUs) to be delivered subject to specific values of QoS parameters, which are defined through a traffic specification (TSPEC) element. The QSTA is responsible for distributing the granted TXOP to its active TSs. During the CP, QSTAs can be polled after the medium is detected idle for a time interval that is shorter than any interval used in EDCA, to ensure access priority of HC over the competing QSTAs. Through this poll, the HC can explicitly assign a TXOP to a particular

3559

QSTA for a specific duration of time. During the CFP, QSTAs do not attempt to get medium access on their own, but rather wait for the HC to assign TXOPs to them. The major responsibility of the HC is to assign TXOPs to STAs in such a way that the negotiated traffic and QoS characteristics of all TSs (as expressed by their TSPECs) are satisfied. The main parameters of a TSPEC are briefly described below [7] (related parameters are grouped together): Minimum (mR) /Mean (ρ) /Peak (PR) Data Rate: Minimum/Average/Peak bit rate for transfer of the packets, in units of bits per second. Delay Bound (D): Maximum delay allowed for the transport of a packet across the wireless interface (including the queuing delay), in milliseconds. Nominal (L) /Maximum (M) MSDU size: Nominal/Maximum size of the MAC layer payload, in octets. Maximum Burst Size (MBS): Maximum size in octets of the data burst that can be transmitted at the peak data rate. Minimum physical rate (R): Physical bit rate assumed by the HC for transmission time and admission control calculations, in units of bits per second. Minimum (mSI) /Maximum (MSI) Service Interval: Minimum/Maximum allowed interval of time, in milliseconds, between the start of two successful transmissions of the particular TS. As aforementioned, a central component of HC for enabling parameterized QoS provision is the traffic scheduler; an entity that basically decides how TXOPs should be efficiently assigned. Note that the traffic scheduler is not part of the 802.11e standard and can thus serve as a product differentiator that should be carefully designed and implemented, since its operation can significantly influence the overall QoS provision. In the open technical literature, only a limited number of 802.11e traffic schedulers have been proposed so far (e.g., [2], [7], [9], [10]) and this paper partially aims at filling this gap. The draft amendment of IEEE 802.11e [2] includes an example scheduling algorithm, referred to as the Simple Scheduler, to provide a reference for future, more sophisticated solutions. The idea of this algorithm is to schedule fixed batches of TXOPs at constant time intervals. Each batch contains one fixed length TXOP per QSTA, based on the mean data rates as declared in the respective TSPECs. With this discipline the Simple Scheduler respects the mean data rates of all TSs and performs well when the incoming traffic load does not deviate from its mean declared value (e.g., constant bit rate traffic). On the other hand, its performance deteriorates significantly when it comes to bursty traffic, as it has no means to adjust TXOP assignments to traffic changes. Identifying the weaknesses of the Simple Scheduler, the scheduling algorithm proposed in [7] provides improved flexibility by allowing the HC to poll each QSTA at variable intervals, assigning variable length TXOPs. The algorithm is referred to as “Scheduling based on Estimated Transmission Times - Earliest Due Date” (SETTEDD), indicating that TXOP assignments are based on earliest deadlines, to reduce transmission delay and packet losses due to expiration. SETT-EDD is a flexible and dynamic scheduler, but it lacks an efficient mechanism for calculating the exact required TXOP duration for each QSTA transmission. Each TXOP duration is estimated based on the mean data rate

3560

IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 5, NO. 12, DECEMBER 2006

of each TS and the time interval between two successive transmissions. In [8] SETT-EDD is extended to optimize the periodic traffic scheduling and include an advanced admission control scheme. The limited capabilities of existing schedulers, together with the advanced scheduling features of HCCA, leave enough space for new proposals able to provide improved performance under a wide range of traffic conditions and QoS requirements.

QS i(x)=5

QS i(x+1)=3

QS i (x+2)=6

TXOPi(x)

TXOPi(x+1)

TXOP i(x+2)

TDi(x)=4

TDi(x+1)=5

TDi(x+2)=3