AGAPE

[Prev] [ToC] [Next]

Experimental Results and Evaluation

We have carried out a number of measurements to evaluate the functioning of AGAPE and to quantify the AGAPE support service overhead. AGAPE introduces different forms of overhead depending on the different AGAPE group and communication management functions involved, from the joining of new members, to members roaming into new localities, to context-dependent view propagation, to on-line advertisement message dissemination, to communication binding management. Moreover, AGAPE group management provides only probabilistic guarantees in group view consistency. This require analizing whether the accuracy degree of our solution fits to emergency rescue domains.

We here show some experimental results that permit to evaluate important factors in collaboration management, i.e. memory, battery degradation and the accuracy degree.

Memory Requirements

We have measured the memory space required for one ME to support advertisement and view management. This is particualrly relevant in our scenario as ME entities are typically characterized by the lack of computational resources and, in particular the lack of memory.

With regard to advertisement management, the main factor that impacts on memory occupation is the table that associates any available AGAPE entity with an entry containing its GID/PID, its role (ME or LME) and its IP address. Each table entry is 21 bytes in size. Context-dependent views can also require a high memory capacity. The memory occupation is determined by the list that stores the views. In particular, in our measurements each entry in the list has an average size of 520 bytes.

In particular, we have measured the memory required by one ME placed in a locality constituted by a variable number of group member entities (5, 10, 15 and 20 members). Similar considerations apply to LMEs. Table 1 shows the memory usage for advertisement and view management.

Table 1. Memory occupation for advertisement and view management.

The measurements show that the memory required both for advertisement and view management linearly grows with the number of members into the locality. However, AGAPE requires a limited amount of memory to properly work. As a consequence, the AGAPE group management support fits devices with limited amounts of memory such as PDAs.

Battery Degradation

Battery degradation depends on several factors, such as group maintenance overhead, group member interaction frequency, and communication costs between the Holter and the elder device to mention a few. Group view dissemination is the AGAPE-related factor that mainly affects battery functioning. VMS requires continuous IEEE 802.11 connections (that introduce typically high energy costs) to disseminate view at regular times. Figure 1 shows that battery life is inversely proportional to the time between consecutive view dissemination. In AGAPE it is possible to change view dissemination rate. Let us note that a low rate, e.g. every minute, preserve batteries but requires an high amount of time for a new member to be included in the group view.

Figure 1. Relation between battery degradation and view dissemination rate

Accuracy Degree

To evaluate the group view accuracy level provided by the AGAPE group membership support we define the Accuracy Degree (Ad) parameter as the fraction of currently connected co-located group members that are included into the current group view. In particular, we here present the variation of the value of the Ad parameter during the time.

Due to the difficulty of deploying a large Mobile Ad-Hoc network, we have simulated the deployment of the emergency rescue prototype over a 4-hops wide locality composed of 1 LME and 49 MEs belonging to the same group. The simulator we use is ns-2 with the Monarch Project wireless and mobile extensions [21]. It provides the implementation of the wireless MAC, based on the Lucent WaveLAN IEEE 802.11 product, with a 2Mbps transmission rate and a nominal range of 250m. We adopt the two-ray ground reflection model as the radio propagation model. We simulate the locality in a square area of 1Km2. The movement pattern is defined by the random way-point model by exploiting the ns-2 setdest tool.

To evaluate Ad we have constantly traced the group view of LME1. For each simulation time, we have calculated the fraction of group members in the locality that are correctly included into the group view. Simulation shows that most of the time the Ad parameter has a value of 1 and that its average value is 0.97. This means that our group management service suits well also for application domains requiring high degree of accuracy in membership management.

[Prev] [ToC] [Next]