Towards Example-Guided Network Synthesis
Towards Example-Guided Network Synthesis
复制标题
迈向示例引导的网络综合
DOI：
--
复制
发表时间：
2018
期刊：
影响因子：
--
通讯作者：
B. T. Loo
中科院分区：
文献类型：
--
作者：
Haoxian Chen;Anduo Wang;B. T. Loo
近年来，网络领域特定语言（DSL）激增。这些语言使我们能够利用这些网络的可编程性，同时仍然通过 DSL 的验证和分析来提供正确性保证。然而，这些 DSL 都没有得到广泛采用。首先，这些新语言需要操作员的学习曲线，而这些操作员可能不是经过培训的程序员。其次，这些新的 SDN 应用有时依赖于传统网络中的功能，而这些功能无法轻松迁移或分析。
为了应对这些挑战，我们提出了 Facon，这是一种新工具，使我们能够根据输入/输出示例自动生成任意 DSL 中的程序。由于输入/输出示例适用于任何网络协议，因此这种方法可以推广，从而使我们能够将遗留网络迁移到新的 DSL，或将一种 DSL 转换为另一种 DSL。作为初步可行性研究，我们将 Facon 应用于一系列基于声明性网络的逻辑网络 DSL。 Facon 以操作员提供的或从遗留系统观察到的输入/输出示例作为示例。 Facon 然后自动生成忠实于这些示例的声明性网络程序。我们提出了一种有效的搜索算法，它利用声明性网络中的句法约束来修剪搜索空间，并利用语义作为启发式来指导搜索方向。我们的初步结果是有希望的。 Facon 成功地综合了声明性网络程序，其规模超出了以前的逻辑程序综合工具的处理能力。
In recent years, there has been a proliferation in network domain-specific languages (DSL). These languages enable us to exploit the programmability of these networks, while still providing correctness guarantees through verification and analysis of DSLs. However, none of these DSLs have received widespread adoption. First these new languages require a learning curve among operators who may not be trained programmers. Second, these new SDN applications sometimes rely on functionality in legacy networks that cannot be easily migrated or analyzed.
To address these challenges, we propose Facon, a new tool that enables us to automatically generate programs in arbitrary DSLs, based on input/output examples. Since input/output examples applies to any network protocols, this approach can be generalized, hence enabling us to migrate legacy networks to new DSLs, or to transform one DSL to another. As an initial feasibility study, we apply Facon to a family of logic-based network DSLs based on declarative networking. Facon takes as examples input/output examples either provided by an operator or observed from a legacy system. Facon then automatically generates a declarative networking program faithful to these examples. We propose an efficient search algorithm that exploits syntactic constraints in declarative networking to prune the search space, and semantics as heuristics to guide the search direction. Our initial results are promising. Facon successfully synthesizes declarative networking programs at a scale beyond previous logic program synthesis tools can handle.
基于约束的数据记录程序综合
DOI：
10.1007/978-3-319-66158-2_44
发表时间：
2017
期刊：
International Conference on Principles and Practice of Constraint Programming
影响因子：
--
作者：
Albarghouthi, Aws;Koutris, Paraschos;Naik, Mayur;Smith, Calvin
通讯作者：
Smith, Calvin