Dataflow systems have an increasing need to support a wide range of tasks in data-centric applications using latest techniques such as machine learning. These tasks often involve custom functions with complex internal states. Consequently, users need enhanced debugging support to understand runtime behaviors and investigate internal states of dataflows. Traditional forward debuggers allow users to follow the chronological order of operations in an execution. Therefore, a user cannot easily identify a past runtime behavior after an unexpected result is produced. In this paper, we present a novel time-travel debugging paradigm called IcedTea, which supports reverse debugging. In particular, in a dataflow's execution, which is inherently distributed across multiple operators, the user can periodically interact with the job and retrieve the global states of the operators. After the execution, the system allows the user to roll back the dataflow state to any past interactions. The user can use step instructions to repeat the past execution to understand how data was processed in the original execution. We give a full specification of this powerful paradigm, study how to reduce its runtime overhead and develop techniques to support debugging instructions responsively. Our experiments on real-world datasets and workflows show that IcedTea can support responsive time-travel debugging with low time and space overhead.

Domain experts play an important role in data science, as their knowledge can unlock valuable insights from data. As they often lack technical skills required to analyze data, they need collaborations with technical experts. In these joint efforts, productive collaborations are critical not only in the phase of constructing a data science task, but more importantly, during the execution of a task. This need stems from the inherent complexity of data science, which often involves user-defined functions or machine-learning operations. Consequently, collaborators want various interactions during runtime, such as pausing/resuming the execution, inspecting an operator's state, and modifying an operator's logic. To achieve the goal, in the past few years we have been developing an open-source system called Texera to support collaborative data analytics using GUI-based workflows as cloud services. In this paper, we present a holistic view of several important design principles we followed in the design and implementation of the system. We focus on different methods of sending messages to running workers, how these methods are adopted to support various runtime interactions from users, and their trade-offs on both performance and consistency. These principles enable Texera to provide powerful user interactions during a workflow execution to facilitate efficient collaborations in data analytics.

Collaborative data analytics is becoming increasingly important due to the higher complexity of data science, more diverse skills from different disciplines, more common asynchronous schedules of team members, and the global trend of working remotely. In this demo we will show how Texera supports this emerging computing paradigm to achieve high productivity among collaborators with various backgrounds. Based on our active joint projects on the system, we use a scenario of social media analysis to show how a data science task can be conducted on a user friendly yet powerful platform by a multi-disciplinary team including domain scientists with limited coding skills and experienced machine learning experts. We will present how to do collaborative editing of a workflow and collaborative execution of the workflow in Texera. We will focus on data-centric features such as synchronization of operator schemas among the users during the construction phase, and monitoring and controlling the shared runtime during the execution phase.

A computing job in a big data system can take a long time to run, especially for pipelined executions on data streams. Developers often need to change the computing logic of the job such as fixing a loophole in an operator or changing the machine learning model in an operator with a cheaper model to handle a sudden increase of the data-ingestion rate. Recently many systems have started supporting runtime reconfigurations to allow this type of change on the fly without killing and restarting the execution. While the delay in reconfiguration is critical to performance, existing systems use epochs to do runtime reconfigurations, which can cause a long delay. In this paper we develop a new technique called Fries that leverages the emerging availability of fast control messages in many systems, since these messages can be sent without being blocked by data messages. We formally define consistency in runtime reconfigurations, and develop a Fries scheduler with consistency guarantees. The technique not only works for different classes of dataflows, but also works for parallel executions and supports fault tolerance. Our extensive experimental evaluation on clusters show the advantages of this technique compared to epoch-based schedulers.

A long-running analytic task on big data often leaves a developer in the dark without providing valuable feedback about the status of the execution. In addition, a failed job that needs to restart from scratch can waste earlier computing resources. An effective method to address these issues is to allow the developer to debug the task during its execution, which is unfortunately not supported by existing big data solutions. In this paper we develop a system called Amber that supports responsive debugging during the execution of a workflow task. After starting the execution, the developer can pause the job at will, investigate the states of the cluster, modify the job, and resume the computation. She can also set conditional breakpoints to pause the execution when certain conditions are satisfied. In this way, the developer can gain a much better understanding of the run-time behavior of the execution and more easily identify issues in the job or data. Amber is based on the actor model, a distributed computing paradigm that provides concurrent units of computation using actors. We give a full specification of Amber, and implement it on top of the Orleans system. Our experiments show its high performance and usability of debugging on computing clusters.

We are developing Texera, an open source system that allows users to perform data analysis on a computing cluster using a GUI-based workflow. A unique functionality of the system is its support for interactive and responsive debugging on dataflows during their execution, while still being scalable and fault tolerant. In particular, users can pause/resume a workflow, investigate the state of operators, change the behavior of an operator, and set conditional breakpoints. In this way, a user will not feel "in the dark" during the long-running execution of an analytics task, a problem faced by other big data processing frameworks. In this demonstration we show this powerful functionality in Texera.