WebOpt Profiler: Providing performance clarity for Web Optimizations
WebOpt Profiler: Providing performance clarity for Web Optimizations
复制标题
WebOpt Profiler：为 Web 优化提供性能清晰度
DOI：
10.1145/3426746.3434050
复制
发表时间：
2020
期刊：
影响因子：
--
通讯作者：
Benson, Theophilus A.
中科院分区：
文献类型：
--
作者：
Murtaza, Ghulam;Benson, Theophilus A.
尽管网络优化的生态系统不断发展，但学术界和工业界的大多数努力都是孤立地设计和开发的。不幸的是，我们现有的工具集[5-7]未能提供允许对Web QoE优化进行详细系统分析的原语。具体来说，我们没有方法来评估现有的优化如何相互叠加？此外，虽然我们继续推出新的Web优化，但尚不清楚它们在什么情况下会提高性能。理解Web优化是一项复杂的任务，因为网页加载过程是复杂的，并且在对象内部和对象之间具有并行和串行处理的任意组合。幸运的是，已经做了大量的工作[5，7，10]来理解网页加载过程。这些工作在现代浏览器的页面加载处理时间轴内将PLT划分为广泛的阶段（联网、处理、布局和绘制）。跨各个阶段的页面加载时间的这种划分有助于研究人员和开发人员了解页面结构，网络，服务器或客户端条件的变化组合如何影响这些阶段粒度的页面加载时间。无论额外的分辨率如何，将Web优化的影响与环境对页面加载过程的影响一分为二是非常具有挑战性的。因此，为了比较Web优化的性能，开发人员要么被迫在一致的环境中艰苦地重新实现优化[5，7]，要么采用冗长且经常出错的过程来建模整个页面加载堆栈，以分析性地讨论性能优势[10]。
Despite the growing ecosystem of web optimizations, most efforts, both within academia and industry, are designed and developed in isolation. Unfortunately, our existing toolset [5–7] fails to provide primitives that allow a detailed systematic analysis of web QoE optimizations. Specifically, we do not have methods to evaluate how existing optimizations stack up against each other? Moreover, while we continue to churn out new web optimizations, it’s unclear under what circumstances they improve performance. Understanding web optimizations is a complex undertaking because the web page load process is intricate and has arbitrary combinations of parallel and serial processing within and across objects. Fortunately, there has been extensive work [5, 7, 10] done to understand the webpage load process. These works divide the PLT into broad stages (Networking, Processing, Layout, and Painting) within the modern browser’s page load processing timeline. This division of page load time across various stages helps the researchers and developers understand how a combination of changes in the structure of the page, network, server, or client conditions impact the page load time at the granularity of these stages. Regardless of additional resolution, it is exceptionally challenging to bisect the impact of web optimization from the environment’s effects on the page load process. Consequently, to compare the performance of web optimizations, the developers are either forced to arduously re-implement optimizations in a consistent environment [5, 7] or adopt the lengthy and often error-prone process of modeling the entire page loading stack to argue about performance benefits analytically [10].
DOI：
10.1145/3341617.3326142
发表时间：
2019-06
期刊：
Proceedings of the ACM on Measurement and Analysis of Computing Systems
影响因子：
--
作者：
Behnam Pourghassemi;A. A. Sani-A.;Aparna Chandramowlishwaran
通讯作者：
Behnam Pourghassemi;A. A. Sani-A.;Aparna Chandramowlishwaran
科兹
影响因子：
22.7
作者：
Charlie Curtsinger;E. Berger
通讯作者：
E. Berger