Information Technology Reference
In-Depth Information
encing severe interference or fading. Therefore,
the performance of these MAC protocols is not
optimized. There are also some proposals such as
(Won, Youn, Ali, Sharif, & Deogun, 2005; Voigt,
Osterlind, & Dunkels, 2008) using the RSSI
value to optimize multi-channel communication.
However, RSSI only characterizes received signal
energy, which does not capture the link charac-
teristics such as channel quality, reliability, and
space and time coherence as our solution does.
It has been shown that optimal performance can
be obtained with EDF policy for a single switch
(Chiussi & Sivaraman, 1998), and certain EDF
techniques can achieve better performance than
those using GPS (Georgiadis, Guerin, Peris, &
Sivarajan, 1996). Yet, the implementation of an
EDF server is more complicated than that of a GPS
one, and proper techniques are needed to make
the cost of such a server affordable in practice.
Several hybrid schedulers that combine EDF
and WFQ (such as (Gidlund & Wang, 2009)
and (Wongthavarawat & Ganz, 2003)) are also
proposed for IEEE 802.16 (also called WiMax)
networks. They apply only either EDF or WFQ but
not both to a flow, depending on the traffic QoS
requirement. For example, in (Wongthavarawat
& Ganz, 2003) EDF is only applied to real-time
Polling Services (rtPS) (for real-time variant bit
rate flows) while WFQ is only applied to the
non-real-time Polling Services (nrtPS) (for non-
real-time flows that require better than best effort
service, e.g., bandwidth-intensive file transfer).
Similarly, EDF is only applied to rtPS while WFQ
is only applied to nrtPS and Best Effort (BE)
service (for best effort traffic such as HTTP with
no QoS requirements). These hybrid schedulers
cannot satisfy the need of vital sign traffic, which
requires both class-based service depending on
patient's condition and the flow-based service to
guarantee the e2e delay requirements, as detailed
in Sect. 4.4.
Packet Scheduling Algorithms
In data networks, packet scheduling has been
used to ensure QoS as a way to control packet
delays. Of all the scheduling techniques, two
classes of scheduling are of great importance in
data networks: Generalized Processor Sharing
(GPS) (Parekh & Gallager, 1993; Parekh & Gal-
lagher, 1994) (also called Weighted Fair Queuing
(WFQ) (Demers, Keshav, & Shenker, 1989)) and
Earliest Deadline First (EDF) scheduling (Chiussi
& Sivaraman, 1998). GPS schedules packets of
each flow with guaranteed minimum bandwidth
according to specified weights. It has been shown
that e2e delay requirements can be mapped into
a bandwidth allocation problem by appropriate
admission control (Parekh & Gallagher, 1994).
It has also been shown that the close coupling
between delay and rate under GPS in deterministic
delay bounds leads to sub-optimal performance
and decreased network utilizations (Georgiadis,
Guerin, Peris, & Sivarajan, 1996).
Conversely, an EDF scheduler assigns “dead-
lines” to packets arriving at the scheduler and
then serves the packets in the ascending order
of their assigned deadlines. Specifically, every
time a packet arrives at one of the queues it is as-
signed a deadline equal to its arrival time plus the
maximum tolerable queuing delay of the packets.
Every packet needs to be sorted according to its
deadline upon its arrival at the node; the packet
with the least deadline is then served first.
PROPOSED NETWORK SCENARIO
Our scenario is assumed to be a network in the
hospital with large enough triage area and limited
number of base stations to collect vital sign data;
in such a scenario multihop communication would
be beneficial to improve the connectivity and to
accommodate more traffic. As shown in Figure
1(a), sensor nodes deployed on a patient form
a Body Area Network (BAN), which monitors,
collects, and pre-processes physiological data of
Search WWH ::
Custom Search