... | ... | @@ -27,39 +27,39 @@ The table describes the activities performed within each experimental unit for a |
|
|
|
|
|
| **Tasks** | **Experimental unit 1 (ph)** | **Experimental unit 2 (ph)** | **Experimental unit 3 (ph)** |
|
|
|
| :--------: | :--------: | :--------: | :--------: |
|
|
|
| Coordination logic | 3 | 100 | 24 |
|
|
|
| Coordination logic | 0 | 100 | 24 |
|
|
|
| Prosumer services | 3,5 | 6 | 24 |
|
|
|
| **Total** | **6,5** | **106 (99,5 saved)** | **48 (41,5 saved)** |
|
|
|
| **Total** | **3,5** | **106 (102,5 saved)** | **48 (44,5 saved)** |
|
|
|
|
|
|
For each experimental unit, the table reports the ph employed to carry out the experimental tasks together with the total amounts of ph. In particular, the total amounts for the experimental units 2 and 3 highlight in brackets the ph saved by using the **CHOReVOLUTION approach**. Specifically, the general-purpose enterprise-oriented approach took more than fifthteen times longer than the **CHOReVOLUTION approach**, whereas the domain-specific system integration platform took more than six times longer.
|
|
|
For each experimental unit, the table reports the ph employed to carry out the experimental tasks together with the total amounts of ph. In particular, the total amounts for the experimental units 2 and 3 highlight in brackets the ph saved by using the **CHOReVOLUTION approach**. Specifically, the general-purpose enterprise-oriented approach took more than twenty-nine times longer than the **CHOReVOLUTION approach**, whereas the domain-specific system integration platform took more than twelve times longer.
|
|
|
|
|
|
**-Hypothesis 2-** We found that the **CHOReVOLUTION** approach provides a meaningful decrease of the time required to maintain the SMT choreography-based system. In the maintenance phase, a different service is selected to play the role of the **Parking** participant, hence leading to a service substitution. The selected service has a different interface with respected to the one required by the choreography specification. In this scenarios, the **CHOReVOLUTION** approach is able to automatically generate additional software entities called Adapters that handle interfaces mismatches. Thus, the experiment tasks considered in this phase, beyond the coordination logic and the prosumer services, include also the experimental task concerning the adaptation logic. In particular, the **CHOReVOLUTION** approach provides automatic support to the generation of the Adapters, whereas the other approaches require a manual implementation or a manual customization.
|
|
|
|
|
|
| **Tasks** | **Experimental unit 1 (ph)** | **Experimental unit 2 (ph)** | **Experimental unit 3 (ph)** |
|
|
|
| :--------: | :--------: | :--------: | :--------: |
|
|
|
| Coordination logic | 0,3 | 8 | 6 |
|
|
|
| Prosumer services | 0,2 | 1 | 1 |
|
|
|
| Adaptation logic | 0,5 | 2 | 1 |
|
|
|
| **Total** | **1** | **11 (10 saved)** | **8 (7 saved)** |
|
|
|
| Coordination logic | 0 | 8 | 3 |
|
|
|
| Prosumer services | 0,2 | 1 | 4 |
|
|
|
| Adaptation logic | 0,5 | 1 | 1,5 |
|
|
|
| **Total** | **0,7** | **10 (9,3 saved)** | **8,5 (7,8 saved)** |
|
|
|
|
|
|
For each experimental unit, the table reports the ph employed to carry out the experimental tasks together with the total amounts of ph. The total amounts for the experimental units 2 and 3 highlight in brackets the ph saved by using the **CHOReVOLUTION** approach. The general-purpose enterprise-oriented approach took ten times longer than the **CHOReVOLUTION** approach, whereas the domain-specific system integration platform took seven times longer.
|
|
|
For each experimental unit, the table reports the ph employed to carry out the experimental tasks together with the total amounts of ph. The total amounts for the experimental units 2 and 3 highlight in brackets the ph saved by using the **CHOReVOLUTION** approach. The general-purpose enterprise-oriented approach took more than thirteen times longer than the **CHOReVOLUTION** approach, whereas the domain-specific system integration platform took eleven times longer.
|
|
|
|
|
|
**-Hypothesis 3-** We found that the **CHOReVOLUTION** approach significantly reduces the time required to evolve the SMT choreography-based system. In the evolution phase, the choreography is modified by removing the task **Get Parking Information** in the left-most parallel branch and the conditional branch concerning the public transportation mode.
|
|
|
|
|
|
| **Tasks** | **Experimental unit 1 (ph)** | **Experimental unit 2 (ph)** | **Experimental unit 3 (ph)** |
|
|
|
| :--------: | :--------: | :--------: | :--------: |
|
|
|
| Coordination logic | 3 | 24 | 20 |
|
|
|
| Prosumer services | 1 | 6 | 4 |
|
|
|
| **Total** | **4** | **30 (26 saved)** | **24 (20 saved)** |
|
|
|
| Coordination logic | 0 | 25 | 6 |
|
|
|
| Prosumer services | 1 | 2 | 8 |
|
|
|
| **Total** | **1** | **27 (26 saved)** | **14 (13 saved)** |
|
|
|
|
|
|
For each experimental unit, the table reports the ph employed to carry out the experimental tasks together with the total amounts of ph. The general-purpose enterprise-oriented approach took more than six times longer than the **CHOReVOLUTION** approach, whereas the domain-specific system integration platform took five times longer. In this phase the choreography has been modified by removing parts of it. This allowed all the three experimental units to leverage on code reuse. It is worth to note that time saving obtained in this phase is due to the high support to automation provided by the **CHOReVOLUTION** approach with respect to the other two approaches that require a manual implementation or customization, although reusing some code.
|
|
|
For each experimental unit, the table reports the ph employed to carry out the experimental tasks together with the total amounts of ph. The general-purpose enterprise-oriented approach took more twenty-six times longer than the **CHOReVOLUTION** approach, whereas the domain-specific system integration platform took thirteen times longer. In this phase the choreography has been modified by removing parts of it. This allowed all the three experimental units to leverage on code reuse. It is worth to note that time saving obtained in this phase is due to the high support to automation provided by the **CHOReVOLUTION** approach with respect to the other two approaches that require a manual implementation or customization, although reusing some code.
|
|
|
|
|
|
**Overall experiment results**
|
|
|
|
|
|
| **Experimental units** | **Implementation (ph)** | **Maintenance (ph)** | **Evolution (ph)** | **Time saving (ph)** |
|
|
|
| :--------: | :--------: | :--------: | :--------: | :--------: |
|
|
|
| 1 | 6,5 | 1 | 4 | - |
|
|
|
| 2 | 106 **(99,5 saved)** | 11 **(10 saved)** | 30 **(26 saved)** | **135,5** |
|
|
|
| 3 | 48 **(41,5 saved)** | 8 **(7 saved)** | 24 **(20 saved)** | **68,5** |
|
|
|
| 1 | 3,5 | 0,7 | 1 | - |
|
|
|
| 2 | 106 **(102,5 saved)** | 10 **(9,3 saved)** | 27 **(26 saved)** | **137,8** |
|
|
|
| 3 | 48 **(44,5 saved)** | 8,5 **(7,8 saved)** | 14 **(13 saved)** | **65,3** |
|
|
|
|
|
|
The table summarizes the results of the experiment on the SMT use case by distinguishing the implementation, maintenance, and evolution phases. In particular, the experimental units 2 and 3 highlight in bold the ph saved by using the **CHOReVOLUTION** approach. Beyond the result concerning the hypothesis 1 discussed before, it is worth to note that the **CHOReVOLUTION** approach results in a decrease of the required development time in all the considered phases. The decrease is more significant in the evolution phase, where the changes affect the choreography specification, than in the maintenance phase, where the changes affected the services involved in the choreography-based system. Moreover, the last column contains the total amount of ph saved for each experimental unit. This result together with the amount of ph saved in each experimental unit reveals that the **CHOReVOLUTION** approach has great potential in developing choreography-based systems and the use case got a full benefit from it. |