Simulation Framework for Modeling Large-Scale Distributed Systems

Dobre Ciprian,Cristea Valentin,Iosif C. Legrand

DOI: https://doi.org/10.48550/arXiv.1106.6122

2011-06-30

Abstract:Simulation has become the evaluation method of choice for many areas of distributing computing research. However, most existing simulation packages have several limitations on the size and complexity of the system being modeled. Fine grained simulation of complex systems such as Grids requires high computational effort which can only be obtained by using an underlying distributed architecture. We are proposing a new distributed simulation system that has the advantage of being able to model very complex distributed systems while hiding the computational effort from the end-user.

Distributed, Parallel, and Cluster Computing

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the limitations of existing simulation software when dealing with large - scale distributed systems (such as grid computing systems). Specifically:

1. **Limitations of scale and complexity**: Most existing simulation packages have limitations in the scale and complexity of the simulated systems and cannot effectively handle complex systems containing thousands of computing elements and data streams.
2. **Bottleneck of computing resources**: Simulating complex systems requires huge computing resources, and existing single - machine workstations will quickly exhaust these resources.
3. **Lack of realism**: In order to be verified before being used in actual grid systems, it is necessary to simulate grid resources as realistically as possible, but existing tools are insufficient in this regard.

To solve these problems, the author proposes a new distributed simulation system. The main advantage of this system is that it can simulate very complex distributed systems while hiding the underlying computational complexity, so that users do not need to worry about the allocation of computing resources. By using a distributed architecture, this system can run in parallel on multiple processor units, thereby accelerating the simulation process and improving the efficiency and scalability of the simulation.

### Key point summary:
- **Problem background**: The limitations of computing resources and complexity encountered by existing simulation tools when dealing with large - scale distributed systems.
- **Solution**: Propose a new distributed simulation framework to overcome the limitations of existing tools through distributed computing.
- **Objective**: Achieve efficient and realistic simulation of very complex distributed systems to support the needs of large - scale grid computing research.

### Relevant formulas (if applicable):
Since this paper mainly involves system design and architecture and does not involve specific mathematical or physical formulas. However, when describing performance optimization and scheduling algorithms, some computational models and performance indicators may be involved, for example:

- **Performance value calculation**:
\[
P_i = f(\text{CPU load}, \text{memory availability}, \text{network latency}, \text{node load})
\]
where \( P_i \) represents the performance value of the \( i \) - th simulation agent, and \( f \) is a function that comprehensively considers multiple factors.

- **Shortest path calculation**:
\[
d_{ij} = \frac{P_i + P_j}{2}
\]
where \( d_{ij} \) represents the weight of the edge connecting two nodes \( i \) and \( j \), based on the arithmetic average of their performance values.

These formulas are used to optimize the scheduling algorithm to ensure the efficient execution of simulation tasks in the distributed system.