07.01.2013 Views

Lecture Notes in Computer Science 3472

Lecture Notes in Computer Science 3472

Lecture Notes in Computer Science 3472

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

or (must-test<strong>in</strong>g):<br />

9 Test<strong>in</strong>g Theory for Probabilistic Systems 249<br />

If all paths <strong>in</strong> P�T are successful then all paths <strong>in</strong> Q�T are successful.<br />

For fully probabilistic processes we do not dist<strong>in</strong>guish between may- and musttest<strong>in</strong>g.<br />

There are two approaches to relate P, Q ∈ FPP:<br />

and<br />

”The probability of all successful paths <strong>in</strong> P�T is at most the probability<br />

of all successful paths <strong>in</strong> Q�T .”<br />

The probability of any trace α <strong>in</strong> P�T is at most the probability of α <strong>in</strong><br />

Q�T .<br />

Intuitively speak<strong>in</strong>g, an implementation (Q) should be at least as successful as<br />

the specification (P).<br />

9.7.1 Test<strong>in</strong>g with Non-probabilistic Test Processes<br />

In this section, we present the approach of Christoff [Chr90]. He presents a test<strong>in</strong>g<br />

theory for fully probabilistic processes that is based on reactive, non-probabilistic<br />

test processes without <strong>in</strong>ternal actions. Two fully probabilistic processes P and<br />

Q are related with regard to their trace distributions while runn<strong>in</strong>g <strong>in</strong> parallel<br />

with a non-probabilistic test process. Note that T np,re is a proper subset of T np ,<br />

the set of all non-probabilistic test processes (compare Def<strong>in</strong>ition 9.7, page 245).<br />

Def<strong>in</strong>ition 9.14. [Chr90] Let P, Q ∈ FPP.<br />

• P ⊑ tr CH<br />

Q iff ∀ T ∈Tnp,re seq , α ∈ Act ∗ :<br />

Pr trace<br />

trace<br />

P�T (α) � PrQ�T (α).<br />

• P ⊑ wte<br />

CH Q iff ∀ T ∈Tnp,re , α ∈ Act ∗ :<br />

�<br />

a∈Act<br />

�<br />

trace PrP�T (α a) � a∈Act<br />

• P ⊑ ste<br />

CH Q iff ∀ T ∈Tnp,re , α ∈ Act ∗ :<br />

Pr trace<br />

trace<br />

P�T (α) � PrQ�T (α).<br />

Recall that T np,re<br />

seq<br />

than or equal to ⊑wte CH<br />

trace PrQ�T (α a).<br />

⊓⊔<br />

is coarser<br />

can be characterized by<br />

conta<strong>in</strong>s only ”sequential” test processes, so ⊑ tr CH<br />

and ⊑ste<br />

CH . It turns out that ⊑tr CH<br />

trace <strong>in</strong>clusion (hence, the superscript ”tr”), i.e.<br />

P ⊑ tr CH Q iff {β ∈ Act∗ | Pr trace<br />

P<br />

(β) > 0} ⊆{β ∈ Act ∗ | Pr trace<br />

Q (β) > 0}.<br />

The probabilistic <strong>in</strong>formation gets lost due to the restrictions of the test pro-<br />

compares the sum of the trace probabilities whereas the<br />

cesses. S<strong>in</strong>ce ⊑ wte<br />

CH

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!