Query languages for big data analysis provide user extensibility through a mechanism of user-defined operators (UDOs). These operators allow programmers to write proprietary functionalities on top of a relational query skeleton. However, achieving effective query optimization for such languages is extremely challenging since the optimizer needs to understand data dependencies induced by UDOs. SCOPE, the query language from Microsoft, allows for hand coded declarations of UDO data dependencies. Unfortunately, most programmers avoid using this facility since writing and maintaining the declarations is tedious and error-prone. In this work, we designed and implemented two sound and robust static analyses for computing UDO data dependencies. The analyses can detect what columns of an input table are never used or pass-through a UDO unchanged. This information can be used to significantly improve execution of SCOPE scripts. We evaluate our analyses on thousands of real-world queries and show we can catch many unused and pass-through columns automatically without relying on any manually provided declarations.

In this paper we present the design and implementation of a distributed, whole-program static analysis framework that is designed to scale with the size of the input. Our approach is based on the actor programming model and is deployed in the cloud. Our reliance on a cloud cluster provides a degree of elasticity for CPU, memory, and storage resources. To demonstrate the potential of our technique, we show how a typical call graph analysis can be implemented in a distributed setting. The vision that motivates this work is that every large-scale software repository such as GitHub, BitBucket, or Visual Studio Online will be able to perform static analysis on a large scale. We experimentally validate our implementation of the distributed call graph analysis using a combination of both synthetic and real benchmarks. To show scalability, we demonstrate how the analysis presented in this paper is able to handle inputs that are almost 10 million lines of code (LOC) in size, without running out of memory. Our results show that the analysis scales well in terms of memory pressure independently of the input size, as we add more virtual machines (VMs). As the number of worker VMs increases, we observe that the analysis time generally improves as well. Lastly, we demonstrate that querying the results can be performed with a median latency of 15 ms.