X hits on this document





4 / 5

always requires that a set of programs in MANE need them.

files be available since

If the path-loss matrix is not pre-computed, the MANE server will use the mobility traces (contents of the files) and a propagation model, e.g., the FSPL model, to compute the path losses for every network packet it receives. Then it uses the computed result to determine

the packet error rate (sending/receiving between

for that those two

particular packet nodes), which affect

its decision to forward or drop the packet.

For every packet generated by a test node, it requires n-1 computation of path losses to determine link connectivity in the emulated wireless MANET environment. Hence, the path-loss computation is computationally intensive and heavily depends upon the number of test nodes in a given topology. Therefore, it may not be computationally prac- tical to determine the wireless link connectivity in real time if the number of test nodes drastically increases (e.g., an order of magnitude) for a realistic scenario or the num- ber of packets being generated by test nodes substantially increases (e.g., a fast data rate network) for a practical data

rate communication. Using the

option will help accom-

modate the scalability required in the MANE emulation by alleviating two challenges for the MANE server: (1) mi- nimizing computation and (2) speeding up data packet forwarding in near real time. The time it takes to compute link connectivity is sped up by making the path-loss in- formation readily available and by eliminating the duplica- tion of computation for packets of same sources that are sent during the same topology (in very close time togeth- er) where locations of test nodes do not change. Minimiz- ing computation will clearly enhance the ability of the MANE server to quickly decide whether network packets should be forwarded (in near real time) to emulate link connectivity between test nodes.

Evaluating the performance of the MANE system with and

without the

option presents quite a

challenge due to the limited amount of time available to accurately and efficiently test, compare, and analyze re- sults. The basis of evaluation will be determining if there is a level of accuracy gained by using the pre-computed path-loss matrix instead of the free space path loss model and if there is any system degradation (loss of performance in processing link connectivity) due to heavy computation requirement. The performance of the MANE system de- pends on a number of factors such as processors used by the MANE server (where the computation of link connec-

tivity occurs), capacity of the local area network (LAN) connecting MANE server and test nodes (for emulating wireless interfaces of test nodes), and the number of test nodes emulated in a MANET scenario. A simple way to test system degradation is to use a large number of test nodes as discussed above. Using a large number of test nodes is restricted by the current hardware implementation (which is limited to 48 test nodes). ARL has resorted to use virtualization technology to implement a large number of test nodes (virtual machines as virtual nodes). There- fore, efforts have been concentrated to implement and tune hardware and software to support emulation of large num- ber of virtual test nodes. Future tests will be conducted to determine whether using this method (pre-computed path loss) on a large scale topology (greater than 100 nodes) will sustain system performance and link connectivity for properly emulating a MANET environment.

The pre-computed option is only usable if the GPS infor- mation of the test nodes are made available prior to real time execution of the MANE server. One way to resolve the computation of the path-loss matrix in real time is to use a multi-core processor system for the MANE server and run on a separate processor as a thread or a separate process and compute and update the path-loss matrix table only for nodes having positional changes in a subsequent network topology. The cost of computing the path-loss in the aforementioned approach is sped up for packets with the same source and destination and being transmitted close in time. This minimizes the number of updates being

made to the path-loss matrix.

The main obstacle of implementing the computing path loss using TIREM in real time is the unavailability of the TIREM source code or the compiled TIREM library for the same executing environment of MANE software (which is Linux environment at kernel 2.6 level). Therefore, ARL was able to successfully employ only the Windows version of the TIREM library to compute the path loss.

The pre-processing technique for using and integrating TIREM into the MANE software system provides more precision in determining the link connectivity of an emulated MANET environment and expedites the decision

of forwarding network process of emulating anticipation, using the

packets, which is the fundamental a MANET environment. With pre-computed path loss option in




















to 1000 nodes. The fast decision traffic also provides more fidelity of high data rates. Considerations

of forwarding network in testing network links and plans of using high

Document info
Document views15
Page views15
Page last viewedThu Jan 19 13:17:39 UTC 2017