Information Technology Reference
In-Depth Information
Algorithm 1. Adaptive GPS Algorithm
if
Receive Feedback From the Sink
then
// Heavy Congestion
if
γ
e
2
// Light Congestion
if
γ
e
2
=
and
γ
e
2
=
then
<
and
φ
G
≠0
then
if
δ
e e
2
> ∆
and
δ
e e
e e
2
2
2
> ∆
then
e e
R
Y
φ
=
φ
+
φ
/ 2
;
φ
=
φ
/ 2
φ
=
φ
+
β φ
+
β φ
R
R
G
G
G
G
G
Y Y
R R
end if
if
γ
e
2
φ
= −
(
1
β φ
)
;
φ
= −
(
1
β φ
)
<
and
φ
Y
≠0
and
φ
G
=0
then
Y
Y Y
R
R R
end if
if
δ
e e
φ
=
φ
+
φ
/ 2
;
φ
=
φ
/ 2
2
≤ ∆
and
δ
e e
e e
2
2
> ∆
then
e e
2
R
R
Y
Y
Y
R
Y
end if
// Intermediate Congestion
if
γ
e
2
(
1
)
φ
=
φ
+
β φ
;
φ
= −
β φ
G
G
Y Y
Y
Y Y
end if
if
δ
e e
=
and
γ
e
2
<
and
φ
G
≠0
then
2
> ∆
and
δ
e e
e e
2
2
≤ ∆
then
e e
2
φ
=
φ
+
φ
/ 2
;
φ
=
φ
/ 2
R
Y
Y
Y
G
G
G
φ
=
φ
+
β φ
;
φ
= −
(
1
β φ
)
end if
if
γ
e
2
G
G
R R
R
R R
=
and
γ
e
2
end if
end if
end if
<
and
φ
G
=0
then
if
δ
e e
e e
2
> ∆
then
2
R
φ
=
φ
+
α φ
;
φ
= −
(
1
α φ
)
Y
Y
R R
R
R R
end if
end if
traffic will not be greatly degraded. For these
cases, the coefficients for less important traffic
are adjusted in a coarser way by taking half of the
original value than that for the 'red' traffic in
order to guarantee the service to the 'red' traffic.
In the case that the service to the 'red' traffic is
guaranteed, the coefficients are adjusted using a
smaller step size. In this way, ϕ
R
, ϕ
Y
, and ϕ
G
can
approach their optimal values with iterations.
Similar adjustments are used in other congestion
cases.
Once (ϕ
R
, ϕ
Y
, ϕ
G
) is updated, the scheduler will
re-calculate
virtual finish time
(Parekh & Gallager,
1993) and schedule the transmission based on the
new values. Notice that as only 3 packets (one
from each class) are to be scheduled each time,
updating virtual finish time is quick. Hence, the
transmission disorder problem
(Wang, Fan, He,
& Tafazolli, 2008) can be easily handled.
while limiting the EDF scheduling complexity.
With the EDF algorithm, every packet in a group
needs to be sorted according to its deadline when it
arrives at the node. The complexity for this sorting
operation is
O
(log
k
), where
k
is the number of
packets in the buffer. One original method to reduce
the complexity is to maintain a number of queues
for the flows into the node. Arriving packets are
put into the end of one queue that can guarantee
the delay requirement properly. The complexity
is reduced to
O
(log
N
), where
N
is the number of
flows arriving at the node.
In a heterogeneous environment, such as for
wireless vital sign monitoring networks, there are
many types of data with different requirements
and delay constraints. The maximum tolerable
queuing delay by this data can therefore vary
depending on the temporal correlation of the vital
sign: it is small for vital signs such as ECG read-
ings and larger for temperature readings. Delay
constraint is critical for vital healthcare data. On
the other hand, it is also necessary to guarantee
that the delay insensitive data do not starve, i.e.,
do not suffer from an uncontrolled delay so to
keep certain QoS provisioning.
Flow-Based Scheduling Optimization
With the triple (ϕ
R
, ϕ
Y
, ϕ
G
) tuned, the next step is
to adjust the number of queues within each class
so that the 'bandwidth inefficiency' is minimized
Search WWH ::
Custom Search