Delay-based and QoS-aware packet scheduling for RT and NRT multimedia services in LTE downlink systems
Madi et al. EURASIP Journal on Wireless Communications and
Networking
Delay-based and QoS-aware packet scheduling for RT and NRT multimedia services in LTE downlink systems
Nadim K. M. Madi 0
Zurina Mohd Hanapi 0
Mohamed Othman 0 1
Shamala K. Subramaniam 0
0 Department of Communication Technology and Networks, Universiti Putra Malaysia , 43400 UPM, Serdang, Selangor D.E. , Malaysia
1 Computational Science and Mathematical Physics Lab, Institute of Mathematical Science, Universiti Putra Malaysia , 43400 UPM, Serdang, Selangor D.E. , Malaysia
Guaranteeing Quality of Service (QoS) for heterogeneous traffic is a major challenge in the Fourth Generation (4G) mobile networks. Therein, the absence of sophisticated resources allocation process at the base station jeopardizes QoS in terms of latency data transfer. It has been observed from the literature that low delay bounds might be ensured, however, at the expense of other QoS aspects; for example, throughput and data loss. Therefore, in this article, we propose an effective Delay-based and QoS-Aware Scheduling (DQAS) scheme with a low complexity overhead as an efficient solution for the resource allocation issue in LTE Medium Access Control (MAC) layer. The ultimate aim of DQAS is to minimize delay for Real-Time (RT) traffic while still offering a good level of QoS. Complying with QoS of different traffic types, we effectively analyze the queue buffer of each user flow by developing an algorithm called Efficient Delay Control (EDC) that weights each flow priority in terms of delay. Then, this weight is utilized as a principle for the scheduling decision on the attending flows. Furthermore, the Least Delay Increase (LDI) algorithm is developed to tune the scheduler behavior to maintain a balance between delay and system throughput. Simulation results considering different user mobility levels reveal that DQAS significantly guarantees a low end-to-end delay trend that is independent of increased RT load, and moreover, a reasonable throughput and data drop levels compared to other existing schedulers.
4G LTE networks; Downlink packet scheduling; Resource allocation; Delay; QoS awareness; QoS balancing
1 Introduction
Long Term Evolution (LTE) is recently the most
promising mobile technology which allows various multimedia
applications to be transferred with a high network
capacity and utility [
1
]. LTE employs Orthogonal Frequency–
Division Multiple Access (OFDMA) as a radio access
technology at the downlink channel; this grants more
flexibility by contiguously utilizing portions of the spectrum
to maximize the network performance.
QoS provisioning has been defined as a major
objective in 4G LTE radio access networks. Therein, MAC
layer scheduling, which is a gist function in radio resource
management (RRM) entity of LTE network architecture,
presents an immanent challenging issue that seeks
effective and realistic solutions to conform with the variety of
data traffic evolution.
In channel–aware scheduling [
2–5
], a trivial scheduling
principle is based on the users’ reported channel
information. This may return a reasonable data rate level.
However, in scenarios of multi–traffic types, channel–
aware scheduling concept by itself is not sufficient to
guarantee a good network QoS performance, particularly
on RT applications. According to [6], the maximum
tolerated delay for RT applications is defined to be less
than 0.1 s and 0.3 s for Voice-over-IP (VoIP) and video
flows, respectively; otherwise, traffic QoS is deteriorated.
Commonly, queue–related parameters are adopted beside
channel rate to allow obtaining a delay–awareness trait.
For example, buffer delay with a maximum bound has
been adopted in many works (see for instance [
7–9
])
to assign delay–oriented scheduling weights to
different flows. A straightforward QoS improvement might be
realized, the performance, in this case, is compromised
during high offered load though. The reason here is that
these scheduling rules are designed based on a
singledimension QoS consideration, hence this precludes the
scheduler to tweak its behavior to adopt changes of the
network load states [
10
].
In addition to that, minimum delay is observed to be
guaranteed only over a single traffic type and at the
expense of low QoS on other types. For example, in [
11,
12
], the algorithms are observed to reduce delay only
on burst RT video, this indeed causes a high data loss
ratio though. Basically, these schemes deliberately trigger
a dropping procedure during the network load
congestion states against certain flows in order to alleviate delay
and improve throughput on other flows with good
channel conditions. Such an excessive dropping event usually
leads to a frequent data retransmissions which eventually
deteriorates delay and QoS on delay–sensitive and small
flows like VoIP.
Motivated by the aforementioned scheduling issues, we
remark that maximizing system throughput is not always
the main goal, but rathe (...truncated)