Information Technology Reference
In-Depth Information
10
20
10
15
10
10
10
5
10
0
10
0
10
1
10
2
10
3
10
4
10
5
10
6
AS link rank
Fig. 4.
Load of AS links
0.3. Moreover, as anticipated, with the same tolerance parameter, the cascading
effect under intentional attack is greater than that under random breakdown.
However, due to the automatically restoration of some faulty links, the difference
between these two cases is deflated comparing with previous research [3].
1.01
1
0.99
α
= 0.1
α
= 0.3
α
= 0.5
α
= 0.7
α
= 0.9
0.98
0.97
0.96
0
2
4
6
8
10
12
14
16
18
20
22
Time step
Fig. 5.
Reachability under intentional attack
Fig. 7 and Fig. 8 are average numbers of rerouting messages in every time step
under different conditions.
RM
(
t
)iscalculatedas(
v∈V
c
RM
v
(
t
))
/|V
c
|
.From
this metric we confirm our previous findings that the cascading failures emerge
when the tolerance parameter of links is less than 0.1. And a failed link with
higher trac load can cause more added burden on the routing system.
Fig. 9 and Fig. 10 show distributions of rerouting messages associated with
every core AS under different conditions. In these cases, we put emphasis on
comparing the rerouting messages generated by initial failure and by cascading
failures. So we just consider conditions that
α
is 0.1, 0.5 and 0.9, represented
as low, median and high tolerant AS links. In the case of initial failure,
RM
v
is the number of rerouting messages in the initial time, i.e., the first time step.
In the case of cascading failures,
RM
v
is calculated as (
t∈T
RM
v
(
t
))
/
, i.e.,
the average number of rerouting messages during the following 20 time steps.
From the distributions we can see that most of core ASes encounter dramatically
increased routing messages due to the cascading effect on the routing system.
Especially when
α
is 0.1, the failures tend to affect all the core ASes, causing
a global effect. Table 1 shows the number of rerouting messages that the 95th
|
T
|
