Information Technology Reference
In-Depth Information
SS-S
3
sends Refresh() to SS-R
1
and SS-R
2
with the parameters (False,
) since no
changes have been detected in the corresponding SS-S
3
cycle. At t
7
, SS-R
1
and SS-R
2
note the reception of the Refresh() sent by SS-S
3
. Let us now assume a server failure
in WS
3
(and hence SS-S
3
) at t
34
. At t
9
, SS-R
1
and SS-R
2
find out that they did not
receive Refresh() from SS-S
3
during the last SS-R
2
cycle. If Max-Retry
2
is equal to 1,
SS-R
1
and SS-R
2
conclude that SS-S
3
failed and hence call the
React()
function.
∅
4 Performance Evaluation
We conducted experiments to assess the different parameters that may impact the
performance of the proposed protocol. We used Microsoft Windows Server 2003
(operating system), Microsoft Visual Studio 8 (development kit), UDDI Server, IIS
Server, and SQL Server. We ran our experiments on
Intel(R) processor (1500MHz)
and 512MB of RAM. Soft-state senders and receivers have been developed in C#.
We created twenty (20) receivers and fifty (50) senders, and registered them in UDDI.
Each receiver has ten (10) senders randomly selected among the existing senders. In
the rest of this section, we analyze the relationship between
τ
SSS
values and the
following two parameters:
fault propagation time
and
false faults
.
τ
SSR
/
200000
180000
160000
SS-R Timer=30s
140000
SS-R Timer=60s
120000
SS-R Timer=90s
100000
80000
SS-R Timer=120s
60000
SS-R Timer=150s
40000
SS-R Timer=180s
20000
0
10
20
30
40
50
60
70
80
90
100
Fault Ratio (%)
Fig. 6.
Impact of τ
SSR
on Fault Propagation Time
Fault propagation time
is the first performance parameter we analyze in our study
(Fig. 6). Let us assume that a fault occurred in a sender at time t
1
and has been de-
tected by a receiver at time t
2
. The fault propagation time is equal to t
2
-t
1
(i.e., the
time it took to the receiver to detect a fault in its senders). Fig. 6 compares the aver-
age fault propagation time for various
τ
SSR
timer values. We consider different fault
ratios for each
SSR
timer value. For instance, a fault ratio of 10 means that 10% (1
out of 10) of participants within a composite service failed. We focused on physical
node faults; these are created by physically stopping the services corresponding to
τ
