Information Technology Reference
In-Depth Information
Fig. 7.4 Synthesis statistics (solutions) for the microarray data analysis scenario
a single solution of length 4 and 344 solutions of length less than or equal
to 11. Even in a search depth of 20, a comparatively small number of 3,281
solutions is returned. An additional constraint that enforces the use of a
particular analysis step ( Netto CustomModel ) even leads to 0 (depth 4), 56
(depth 11) and 680 (depth 20) obtained results.
When synthesizing microarray data analysis workflows in the scope of the
scenario described in Section 6.3, the maximum capacity of the solution store
is reached in a search depth of 8 if no constraints are applied. Due to the
complexity of the analysis, there are, however, several workflow variants that
involve more steps, so it is clearly necessary to enable the synthesis to pro-
ceed to greater search depths. Figure 7.4 shows that already the domain
constraints restrain the growth of the amount of solutions effectively, so that
6,855 results are returned in a search depth of 8. Additionally enforcing the
use of a particular benchmark data set ( LoadSpikeInBenchmarkData )and
the GetPubMedAbstracts annotation service tames the amount of obtained
solutions further, so that 385 solutions of length less than or equal to 8
remain.
7.1.2 Performance
The runtime performance of the synthesis algorithm largely depends on the
domain model: The synthesis universe (cf. Section 2.3.1) is constituted by the
(static) service descriptions, more precisely by the directed graph constructed
from their behavioral characterization in terms of input and output data
types. Thus, the number of states in the synthesis universe is exponential
in the number of services in the domain model, an effect that is commonly
known as state explosion [324]. In other words, the more behaviorally similar
tools (i.e., tools with similar inputs or outputs) are available in the domain
 
Search WWH ::




Custom Search