| 1. |
|
|
| 2. |
- Kunz, Georg, et al.
(författare)
-
Expanding the Event Horizon in Parallelized Network Simulations
- 2010
-
Ingår i: 2010 IEEE International Symposium on Modeling, Analysis & Simulation of Computer and Telecommunication Systems (MASCOTS). - : IEEE conference proceedings. - 9781424481811 ; , s. 172-181
-
Konferensbidrag (refereegranskat)abstract
- The simulation models of wireless networks rapidly increase in complexity to accurately model wireless channel characteristics and the properties of advanced transmission technologies. Such detailed models typically lead to a high computational load per simulation event that accumulates to extensive simulation runtimes. Reducing runtimes through parallelization is challenging since it depends on detecting causally independent events that can execute concurrently. Most existing approaches base this detection on lookaheads derived from channel propagation latency or protocol characteristics. In wireless networks, these lookaheads are typically short, causing the potential for parallelization and the achievable speedup to remain small. This paper presents Horizon, which unlocks a substantial portion of a simulation model's workload for parallelization by going beyond the traditional lookahead. We show how to augment discrete events with durations to identify a much larger horizon of independent simulation events and efficiently schedule them on multi-core systems. Our evaluation shows that this approach can significantly cut down the runtime of simulations, in particular for complex and accurate models of wireless networks.
|
|
| 3. |
- Kunz, Georg, et al.
(författare)
-
Extending the OMNeT++ Sequence Chart for Supporting Parallel Simulations in Horizon
- 2012
-
Konferensbidrag (refereegranskat)abstract
- Developing parallel network simulations is a complex task. Besides getting the model right, developers of parallel simulations are striving for an additional design goal: Performance. We argue that developers need an insight into the behavior of a simulation model in order to assess and optimize its parallel performance. Specically, given a parallel simulation model, it is imperative to identify and eliminate performance bottlenecks. To this end, we extend the sequence chart provided by the Eclipse-IDE of OMNeT++ with specic functionality to visualize, analyze, and optimize the performance of parallel simulations in the context of our OMNeT++-based parallel simulation framework Horizon. This extended abstract presents the features and modifications of our code contribution.
|
|
| 4. |
- Kunz, Georg, et al.
(författare)
-
Know Thy Simulation Model : Analyzing Event Interactions for Probabilistic Synchronization in Parallel Simulations
- 2012
-
Ingår i: SIMUTOOLS '12. - : ACM. - 9781450315104 ; , s. 119-128
-
Konferensbidrag (refereegranskat)abstract
- Efficiently scheduling and synchronizing parallel event exe-cution constitutes the fundamental challenge in parallel dis-crete event simulation. Existing synchronization algorithmstypically do not analyze event interactions within the sim-ulation model – mainly to minimize runtime overhead andcomplexity. However, we argue that disregarding event in-teractions results in a lack of insight into thebehaviorofthe simulation model, thereby severely limiting synchroniza-tion efficiency and thus parallel performance. In this pa-per, we present a probabilistic synchronization scheme thatobtains extensiveknowledgeof the simulation behavior atruntime to guide event execution. Specifically, we designthreeheuristicsthat dynamically derive event dependenciesfrom tracing event interactions and decide whether or notto speculatively execute events. Our evaluation shows thatthe proposed probabilistic synchronization scheme consid-erably outperforms traditional conservative and optimisticschemes.
|
|
| 5. |
- Kunz, Georg, et al.
(författare)
-
Multi-Level Parallelism for Time- and Cost-efficient Parallel Discrete Event Simulation on GPUs
- 2012
-
Ingår i: 2012 ACM/IEEE/SCS 26th Workshop on Principles of Advanced and Distributed Simulation (PADS). - : IEEE conference proceedings. - 9780769547145 ; , s. 23-32
-
Konferensbidrag (refereegranskat)abstract
- eveloping complex technical systems requires a systematic exploration of the given design space in order to identify optimal system configurations. However, studying the effects and interactions of even a small number of system parameters often requires an extensive number of simulation runs. This in turn results in excessive runtime demands which severely hamper thorough design space explorations. In this paper, we present a parallel discrete event simulation scheme that enables cost- and time-efficient execution of large scale parameter studies on GPUs. In order to efficiently accommodate the stream-processing paradigm of GPUs, our parallelization scheme exploits two orthogonal levels of parallelism: External parallelism among the inherently independent simulations of a parameter study and internal parallelism among independent events within each individual simulation of a parameter study. Specifically, we design an event aggregation strategy based on external parallelism that generates workloads suitable for GPUs. In addition, we define a pipelined event execution mechanism based on internal parallelism to hide the transfer latencies between host- and GPU-memory. We analyze the performance characteristics of our parallelization scheme by means of a prototype implementation and show a 25-fold performance improvement over purely CPU-based execution.
|
|
| 6. |
- Kunz, Georg, et al.
(författare)
-
Predicting Runtime Performance Bounds of Expanded Parallel Discrete Event Simulations
- 2011
-
Ingår i: Modeling, Analysis & Simulation of Computer and Telecommunication Systems (MASCOTS), 2011 IEEE 19th International Symposium on. - : IEEE conference proceedings. - 9781457704680 ; , s. 359-368
-
Konferensbidrag (refereegranskat)abstract
- Predicting and analyzing runtime performance characteristics is a vital step in the development process of parallel discrete event simulations. For instance, model developers need to identify and eliminate performance bottlenecks within a simulation model in order to derive a model structure that aids parallel execution. Similarly, developers of parallel simulation frameworks require means of assessing the efficiency of the framework. In this paper, we present a performance prediction methodology that computes the best possible performance bound for expanded parallel discrete event simulations in the context of our Horizon simulation framework. The methodology builds upon a linear program which calculates an optimal event execution schedule for a given simulation and a set of CPUs. In order to mitigate the complexity of this NP-complete scheduling problem, we introduce performance optimizations and relaxations of the linear program.
|
|
| 7. |
- Kunz, Georg, et al.
(författare)
-
Runtime Efficient Event Scheduling in Multi-threaded Network Simulation
- 2011
-
Ingår i: Proceedings of the 4th International Workshop on OMNeT++ (OMNeT++'11), Barcelona, Spain. - Brussels, Belgium : ICST. ; , s. 359-366
-
Konferensbidrag (refereegranskat)abstract
- Developing an ecient parallel simulation framework for multiprocessor systems is hard. A primary concern is the considerable amount of parallelization overhead imposed on the event handling routines of the simulator. Besides complex event scheduling algorithms, the main sources of overhead are thread synchronization and locking of shared data. Thus, compared to sequential simulation, the overhead of parallelization may easily outweigh its performance benets. We introduce two ecient event handling schemes based on our parallel-simulation extension Horizon for OMNeT++.First, we present a push-based event handling scheme to minimize the overhead of thread synchronization and locking. Second, we complement this scheme with a novel event scheduling algorithm that signicantly reduces the overhead of parallel event scheduling. Lastly, we prove the correctness of the scheduling algorithm. Our evaluation reveals a total reduction of the event handling overhead of up to 16x.
|
|
| 8. |
- Stoffers, Mirko, et al.
(författare)
-
Large-Scale Network Simulation : Leveraging the Strengths of Modern SMP-based Compute Clusters
- 2014
-
Ingår i: Proceedings of the 7th International ICST Conference on Simulation Tools and Techniques. - : The Institute for Computer Sciences, Social Informatics and Telecommunications Engineering. - 9781631900075 ; , s. -40
-
Konferensbidrag (refereegranskat)abstract
- Parallelization is crucial for ecient execution of large-scale network simulation. Today's computing clusters commonly used for that purpose are built from a large amount of multiprocessor machines. The traditional approach to utilize all CPU cores in such a system is to partition the network and distribute the partitions to the cores. This, however, does not incorporate the presence of shared memory into the design, such that messages between partitions on the same computing node have to be serialized and synchronization becomes more complex. In this paper, we present an approach that combines the shared-memory parallelization scheme Horizon [9] with the standard approach to distributed simulation to leverage the strengths of today's computing clusters. To further reduce the synchronization overhead, we introduce a novel synchronization algorithm that takes domain knowledge into account to reduce the number of synchronization points. In a case study with a UMTS LTE model, we show that both contributions combined enable much higher scalability achieving almost linear speedup when simulating 1,536 LTE cells on 1,536 CPU cores.
|
|
