Information Technology Reference
In-Depth Information
5 Decision Making for Adaptation and Evolution
We use the satisfaction degree calculated using the fuzzy inference system for
the adaptation and evolution decision making. The decision making mechanism
works based on the algorithm we provided in [24]. The algorithm evaluates the
evolution of the service and decides which adaptation strategy to take with
respect to the predefined threshold degree for QoS satisfaction. The two main
decisions are the internal renegotiation in which the changes are compatible
with the service description in the contract and service replacement in which
the changes are incompatible with the existing contract. The former case deals
with the internal contract modification with the same provider and requester
while the earlier case requires the selection of a new service and establishment
of a new contract which can result in a huge loss of time and money.
Having provided such a decision making mechanism allows us to offer a flexible
adaptation mechanism. This is done by identifying threshold to what constitutes
compatible and incompatible . Using satisfaction degree allows us to define the
criticality of a change/violation. Therefore, we are able to understand whether
a violation is critical and it results in a service replacement or the violation is
still acceptable. This way, a slight change from the quality ranges defined in the
contract will not trigger the adaptation. Table 1 shows the result of checking
for compatibility for a possible set of changes. The comparison is between our
fuzzy approach and a traditional non-fuzzy one that works based on the precise
evaluation of the quality ranges in the contract.
Table 1. Comparing the adaptation decisions using fuzzy and non-fuzzy approach
Change
Replacement?
Change
Replacement?
(Non-fuzzy/Fuzzy)
(Non-fuzzy/Fuzzy)
S 1 =(6
,.
90)
Yes/No
S 2 =(7
,.
75)
Yes/Yes
S 3 =(5
,.
85)
No/No
S 4 =(3
,.
70)
Yes/No
S 5 =(2
,.
85)
No/No
S 6 =(2
,.
78)
Yes/No
S 7 =(2
,.
60)
Yes/Yes
S 8 =(3
,.
90)
No/No
S 9 =(7
,.
95)
Yes/Yes
S 10 =(6
,.
50)
Yes/Yes
In the fuzzy approach the replacement is based on the satisfaction degree.
However, in the non-fuzzy approach a service replacement is necessary if any
parameters are violated from the initial range, albeit minor deviation. For ex-
ample in S 1 =(6 ,. 90), changing the response-time to 6s will not result a service
replacement applying the fuzzy approach since it has the satisfaction degree of
almost 83%. While applying a non-fuzzy approach, it is considered a violation
because it does not respect the initial response-time range (2, 5) in the contract.
However, if a change results in a low satisfaction degree, service replacement is
necessary in both approaches as in the case S 10 =(6 ,. 50) which the satisfaction
degree is around 62%.
 
Search WWH ::




Custom Search