Proximity Detection in Distributed Simulation of Wireless Mobile Systems
0 likes596 views
The document discusses proximity detection in distributed simulations of wireless mobile systems, focusing on algorithms to identify which entities are involved in event notifications based on transmission ranges. It evaluates various methods, including the all-pairs (brute force) algorithm and enhancements like grid-based data structures and results caching, showing their performance in managing mobile host interactions. Additionally, it presents experimental results and future work directions, including adaptive grid cell sizes and collision detection improvements.
Proximity Detection in Distributed Simulation of Wireless Mobile Systems
- 1. Gabriele D’Angelo Lorenzo Donatiello Luciano Bononi Michele Bracuto Department of Computer Science University of Bologna Torremolinos, 03/10/2006 Proximity Detection in Distributed Simulation of Wireless Mobile Systems
- 2. Presentation outline Proximity Detection of Wireless Mobile Systems Proximity Detection algorithms + enhancements All-pairs algorithm (brute force) External and internal squares Grid-based data structure Results caching Experimental environment Simulation results Conclusion and future work
- 3. Proximity Detection of Wireless Mobile Systems A common problem of many simulations is to determine what entities are involved in the notification and processing of a given event In a wireless simulation a new frame transmission is translated as the delivery of a transmission event from a sender host to a whole set of potential receiver entities The wireless signals decay with distance depending on propagation model assumptions
- 4. Proximity Detection of Wireless Mobile Systems For performance reasons, it make sense to distribute the transmission event messages only to the subset of neighbor hosts that are reached by the transmission Since usually the hosts are mobile, such subset is very dynamic Under the simulation viewpoint, this problem can be translated in the dynamic identification of the recipient entities of each event message Some work has been done to address the proximity detection over single processor architectures and multi-processors with shared memory but mainly for cost-reasons the distributed simulation architectures are more and more interesting and diffused
- 5. The all-pairs algorithm (brute force) Given a set of wireless devices moving in a simulated scenario, at each timestep, the simulator has to determine the whole set of hosts affected by the ongoing transmissions In a simplistic way this problem can be solved by computing the distance between each pair of nodes. This algorithm is usually referred to as “ all-pairs ” (brute force) and has a complexity of O( n 2 ) per timestep Proposed proximity management solutions: External and internal squares Grid-based data structure Results caching
- 6. External and internal squares 1/2 The all-pairs algorithm computes n(n-1) Euclidean distance values between nodes coordinates By exploiting the “ homogeneous limited horizon distance ” assumption of wireless transmissions, it is possible to reduce the effective cost of the all-pairs algorithm By assuming a 2-D space, in the worst case four comparisons are necessary to determine if a point falls within a square: the evaluation of the Euclidean distance is much more costly!
- 7. External and internal squares 2/2 Hosts allocated outside of the external square can be assumed to be out of the reception range of the central SMH, and can be discarded without any further evaluation The transmission events originated by the SMH can be delivered without any further examination to other SMHs within the inner square. Only the dashed zone has to be further evaluated
- 8. Grid-based data structure An overlay grid divides the simulated space in cells. Each square of the grid (or cell) represents only a small portion of the simulated scenario and has a fixed size The overlay grid is composed by a matrix of separate chaining hash tables Only the grid squares whose area has non zero intersection with the transmission range circle have to be checked In this way the number of checks can be drastically reduced, but the data structure management introduces some overhead
- 9. Results caching The Euclidean distance is a commutative binary operator: given two points x and y we have that d(x,y) = d(y,x) In a centralized simulation-execution architecture it would be possible to reduce the number of Euclidean distance operations to n(n-1)/2 On the other hand, in a distributed simulation, each LP manages only a subset of the SMHs and a global knowledge of the simulated environment is missing A prerequisite to apply the commutative optimization is: in a pair (SMH1, SMH2) both simulated nodes have to be allocated in the same LP. Otherwise the results caching can not be applied
- 10. Simulated Mobile Hosts (SMHs) migration Wireless ad hoc network scenario: (evaluating migration of SMH x) X’s “transmission-event” must be notified to the 4 model entities executed over B After X’s migration, X’s “transmission-event” must be notified to one model entity executed over A A B network delay Physical Execution Units for the simulation model entities of SMHs executed over PEUs A and B Host X executed over PEU A “transmission-event” A B network delay Physical Execution Units for the simulation model entities of SMHs executed over PEUs A and B Same scenario after migration of X from A to B X X X SMH = Simulated Mobile Host PEU = Physical Execution Unit
- 11. The visual effect of the migration based mechanism Migration OFF Migration ON
- 12. Migration & results caching The migration based approach can reduce the communication cost, clustering together the highly interacting entities A side-effect of this mechanism is to allocate the highly interacting entities within the same LP In a wireless simulation the interacting entities are usually quite close in the simulated topology As a consequence, the migration mechanism increases the probability that two interacting mobile hosts are executed within the same LP. Hence, the migration mechanism increases the probability to apply with success the results caching
- 13. Performance evaluation: Ad-Hoc wireless network model Simulated model: 6000 Simulated Mobile Hosts (SMHs) on a torus-shaped bidimensional grid-topology, 10.000x10.000 space units Mobility model: Random Waypoint Model (max speed 10 m/s) uncorrelated SMHs’ mobility Traffic model: ping messages (CBR) by every SMH to all neighbors within the wireless communication range (250 m) Propagation model open space (neighbor-SMHs within detection range)
- 14. Experimental environment and performance evaluation PEU equipped an Intel Dual Xeon Pentium IV 2800 MHz, with 3 GB RAM, Debian GNU/Linux OS with kernel version 2.6.10 Conservative time-stepped simulation: 1000 time-steps Performance evaluation: All-pairs algorithm vs. proposed enhancements The effect of variable grid-side on performances The performance of the caching mechanism The effect of the entities migrations on the proposed algorithms and enhancements
- 15. Simulation results: proximity detection
- 17. Proximity detection: all-pairs with caching ON/OFF
- 18. Proximity detection: enhanced all-pairs
- 19. Proximity detection: all-pairs with migration and caching ON/OFF
- 20. Proximity detection: migration ON/OFF
- 21. Conclusions and future work The proximity detection algorithms are a key part of the simulation of many mobile wireless systems We have analyzed some proposals for proximity detection algorithms being executed in a distributed architecture Some enhancements to the classical algorithms, and specifically tailored data structures have been proposed and evaluated The composition of migration mechanisms and enhanced proximity detection algorithms can improve the simulators performances We plan further investigations of proximity detection schemes (adaptive grid cell-sizes) and related topics (i.e. collision detection)
- 22. Gabriele D’Angelo Lorenzo Donatiello Luciano Bononi Michele Bracuto Department of Computer Science University of Bologna Torremolinos, 03/10/2006 Proximity Detection in Distributed Simulation of Wireless Mobile Systems