Information Technology Reference
In-Depth Information
C
0
C
2
C
1
C
N
x
x
2
=
-
2
d
x
1
x
N
=
-
d
x
0
=0
=
-
N
d
Fig. 1.
The scenario under consideration.
d
is the average inter-vehicle distance.
3.2
The Ben-Arabi Supercomputer
Our model is executed under the Ben-Arabi supercomputer resources, which is placed
in the Scientific Park of Murcia (Spain). The Ben-Arabi system consists of two different
architectures; on the one hand the central node HP Integrity Superdome
SX
2000
with
128
cores of the Intel Itanium-
2
dual-core Montvale (
1
.
6
Ghz,
18
MB of cache L
3
)
processor and
1
.
5
TB of shared memory, called Ben. On the other hand, Arabi is a
cluster consisting of
102
nodes, which offers a total of
816
Intel Xeon Quad-Core E5450
(
3
GHz y
6
MB of cache L
2
) processor cores and a total of
1072
GB of shared memory.
We run our mathematical model within a node of the Arabi cluster environment using
2
,
4
and
8
processors in order to compare the resulting execution times. Les us remark
that we are using a shared memory parallelization technique, so we are not allowed to
combine the use of processors from different nodes.
Next we summarize the technical features of the cluster:
-
Capacity:
9
.
72
Tflops.
-
Processor: Intel Xeon Quad-Core E
5450
.
-
Nodes number:
102
.
-
Processors number:
816
.
-
Processors/Node:
8
.
-
Memory/Node:
32
nodes of
16
GB and
70
of
8
GB.
-
Memory/Core:
3
MB (
6
MB shared among
2
cores).
-
Clock frequency:
3
Ghz.
4
Model Description
We are interested in evaluating the performance of a CCA application for a chain of
N
vehicles when the technology penetration rate is not
100%
. We consider the inter-
vehicle spacing is normally distributed and each vehicle
C
i
,i∈{
1
, ..., N},
moves at
constant velocity
V
i
. Vehicles drive in convoy (see Figure 1), reacting to the first colli-
sion of another car,
C
0
, according to two possible schemes: starting to brake because of
a previously received warning message transmited by a collided vehicle (if the vehicle
is equipped with CCA technology) or starting to decelerate after noticing a reduction
in the speed of the vehicle immediately ahead (if the vehicle under consideration is not
equipped with CCA technology).
With this model the final outcome of a vehicle depends on the outcome of the preced-
ing vehicles. Therefore, the collision model is based on the construction of the following
probability tree. We consider an initial state in which no vehicle has collided. Once the
danger of collision has been detected, the first vehicle in the chain
C
1
(immediately
after the leading one) may collide or stop successfully. From both of these states two
Search WWH ::
Custom Search