UniLoG: A Unified Load Generation Tool
UniLoG: A Unified Load Generation Tool
复制标题
UniLoG：统一的负载生成工具
DOI：
10.1007/978-3-642-28540-0_21
复制
发表时间：
2012
期刊：
影响因子：
--
通讯作者：
A. Kolesnikov
中科院分区：
文献类型：
--
作者：
A. Kolesnikov
用于生成人工（合成）负载的实用程序对于分析网络及其提供的服务的性能和行为非常重要。工业上实现的负载发生器主要是具有非常高的性能和严格的精度要求的专用硬件组件。在研究和学术界中，通常使用主要基于软件的负载生成器，这是因为操作和维护中预期的更高灵活性（例如，由于在网络中容易部署组成负载生成模块、用于特定研究目的的代码定制等）。而真实的操作系统和协议栈的组件可以用于以较低的成本保证现实的负载生成。然而，许多现有的工具专用于特定的建模研究（例如，Guernica [1]沿着其用于Web流量的特定Dweb模型，或Harpoon [2]建模IP流量流），或者专注于在网络中的某个特定接口处生成流量（例如，ITG [3]或Brute [4]被设计为在UDP和TCP服务接口上生成流量）。所提出的解决方案通常不提供足够的灵活性，例如，在要修改基础模型或要使用全新模型的情况下。因此，本文提出了统一的负载生成器UniLoG，它结合了规范和生成网络负载在一个单一的连贯的方法。UniLoG的设计和详细说明的基本原则是通过有限用户行为自动机（乌巴，在第2节中介绍）对抽象负载模型进行形式化描述。2)并且此后使用接口专用适配器将抽象请求映射到具体请求，因为它们被所讨论的真实的接口处的服务提供组件“理解”。分布式UniLoG架构的概述见第2节。3和一个具体的例子，它的实际使用中的QoS研究视频流的演示。4.
Utilities for generating artificial (synthetic) loads are very important for analyses of performance and behavior of networks and their offered services. Load generators implemented by the industry are mainly dedicated hardware components with very high performance and stringent precision requirements. In research and academia, mainly software based load generators are commonly used because of the expected higher flexibility in operation and maintenance (e.g. due to easy deployment of constituent load generating modules in the network, code customizations for a specific research purpose, etc.) while components of real operating systems and protocol stacks can be used to guarantee realistic load generation at lower costs. However, many existing tools are dedicated to a specific modeling study (e.g., Guernica [1] along with its specific Dweb model for Web traffic, or Harpoon [2] modeling IP traffic flows) or are focusing on generating traffic at some specific interface in a network (e.g., ITG [3] or Brute [4] were designed to generate traffic on UDP and TCP service interfaces). The proposed solutions quite often do not provide an adequate flexibility, e.g. in case the underlying model is to be modified or a completely new model is to be used. Therefore, the unified load generator UniLoG is presented in this paper, which combines the specification and generation of network loads in one single coherent approach. The basic principle underlying the design and elaboration of UniLoG is to start with a formal description of an abstract load model by means of a finite user behavior automaton (UBA, introduced in Sec. 2) and thereafter to use interface-specific adapters to map the abstract requests to the concrete requests as they are "understood" by the service providing component at the real interface in question. An overview of the distributed UniLoG architecture is given in Sec. 3 and a concrete example of its practical use in QoS studies for video streaming is demonstrated in Sec. 4.