SGER: Exploratory Research: Interactive Visualization and Control of Mobile Network Simulations

本研究的目的是探索信息可视化方案，以交互地控制、驱动和分析移动网络中自适应资源管理协议的仿真。作为自适应资源管理协议特征的大量系统变量和度量以及对它们进行实时监控的需要导致了数据爆炸，需要使用数据可视化方案来进行有效的分析和优化。将探索两个想法：1.通过适当的仿真变量(系统监控指标、度量参数和仿真系统参数)的可视化交互，交互式地引导仿真来优化多层自适应协议。根据各种动态网络条件下的关键特征(例如，拥塞、故障)对模拟进行表征和分段。目标1旨在使用户成为模拟环境的中心，允许模拟被控制(停止、备份、变量/协议改变)并指向特定的预定目标。系统变量的动态可视化和它们之间的交互是这种交互式探索的基础，更重要的是，用户的直觉和领域专业知识有助于探索(减少)这些变量的大搜索空间。目标2旨在通过关注模拟的关键特征来表征模拟，例如正常、拥塞、瞬时和稳态故障以及恢复条件。检测和跟踪这些特征的健壮方案将允许紧凑地表示、存储和搜索大量的仿真数据。我们的目标是评估这种方法在自动化自适应协议设计和优化方面的应用。两个提出的想法将针对一个特定的问题实例进行评估，该实例正在进行的NSF项目正在进行中，以解决可生存无线接入网络。问题是开发一种可扩展的资源管理方案，该方案将自适应接纳控制和自适应多址接入控制的使用集成在移动网络的分层蜂窝结构中。这种多层次的分析和优化问题提供了一个巨大的仿真场景搜索空间，对任何类型的比较或系统研究都构成了巨大的挑战。早期的工作在很大程度上充其量只限于此类问题的个别组成部分。方法是将其归结为一个多变量数据分析问题，并探索几种交互式多变量可视化方法。这应该有助于理解搜索空间变量之间复杂的空间和时间关系。交互式转向是探索这一空间并达到预期目标的有效手段，而通过提取和参数化关键特征来抽象模拟有助于促进批处理计算运行或网络监控的自动评估。(1)以前没有尝试过将信息可视化应用于移动联网领域。所提出的工作将极大地改变/改进当前的自适应协议开发和基于模拟的分析的方法。然而，这项工作需要在新兴研究领域(交互式转向和特征描述)试验新想法，并且不能保证具体的结果。(2)由于拟议的工作将与新资助的国家科学基金项目相结合，因此需要立即提供支助。预计这将大大提高这两个项目取得的进展。

The objective of this research is to explore information visualization schemes to interactively control, drive and analyze simulations of adaptive resource management protocols in mobile networks. The considerable number of system variables and metrics that characterize adaptive resource management protocols and the need to monitor them in real-time results in a data explosion, necessitating data visualization schemes for efficient analysis and optimization. Two ideas will be explored:1. Interactively steer simulations to optimize multi-layer adaptive protocols, via visualization interactions of appropriate simulation variables (system monitoring metrics, metric paramaters and simulation system parameters).2. Characterize and segment simulations in terms of their critical features (e.g., congestion, failure) under a variety of dynamic network conditions.Objective 1 is aimed at making the user central to the simulation environment, allowing simulations to be conrolled (stopped, backed up, variables/protocols changed) and directed towards certain predefined objectives. Dynamic visualizations of system variables and interactions among them are fundamental to this interactive exploration, and more important, the user's intuition and domain expertise helps explore (reduce) a large search space of these variables. Objective 2 is targeted at characterizing simulations by focusing on critical features of a simulaltion, such as normal, congestion, transient and steady state failure, and recovery conditions. A robust scheme to detect and track these features will permit large amounts of simulation data to be compactly represented, stored and searched. The goal is to evaluate the application of this approach to automating adaptive protocol design and optimization.The two proposed ideas will be evaluated for a specific problem instance that is being addressed by an ongoing NSF project on survivable wireless access networks. The problem is to develop a scaleable resource management scheme that integrates the use of adaptive admission control and adaptive multiple access control within hierarchical cellsite architectures of mobile networks. This multi-layer analysis and optimization problem presents a hugh search space of simulation scenarios, posing significant challenges to any type of comparative or systematic study. Earlier work has largely been restricted to individual components of such problems, at best. The approach is to cast this as a multi-variate data analysis problem, and explore several interactive multi-variate visualization methods. This should help in understanding the complex spatial and temporal relationships among the search space variables. Interactive steering is an efficient means to exploring this space and lead to desired objectives, while abstracting simulations by extracting and parameterizing critical features serves to facilitate automated evaluation of batch siumlation runs or network monitoring.Support by an SGER is being requested for two reasons. (1) The proposed application of information visualization to the mobile networking domain has not previously been attempted. The proposed work shows great promise to dramatically change/improve current approaches to development and simulation-based analysis of adaptive protocols. However, this work requires experimentation with new ideas in emerging research areas (interactive steering and feature characterization), and specific outcomes are not guaranteed. (2) Immediate support is needed since the proposed work is to be conducted in conjucntion with a newly-funded NSF project. It is expected that this will greatly enhance the progress achieved within both projects.