Distributed Dataflow Systems Research Articles

Imperative or Functional Control Flow Handling

Modern data analysis tasks often involve control flow statements, such as the iterations in PageRank and K-means. To achieve scalability, developers usually implement these tasks in distributed dataflow systems, such as Spark and Flink. Designers of such systems have to choose between providing imperative or functional control flow constructs to users. Imperative constructs are easier to use, but functional constructs are easier to compile to an efficient dataflow job. We propose Mitos, a system where control flow is both easy to use and efficient. Mitos relies on an intermediate representation based on the static single assignment form. This allows us to abstract away from specific control flow constructs and treat any imperative control flow uniformly both when building the dataflow job and when coordinating the distributed execution.

Handling Iterations in Distributed Dataflow Systems

Over the past decade, distributed dataflow systems (DDS) have become a standard technology. In these systems, users write programs in restricted dataflow programming models, such as MapReduce, which enable them to scale out program execution to a shared-nothing cluster of machines. Yet, there is no established consensus that prescribes how to extend these programming models to support iterative algorithms. In this survey, we review the research literature and identify how DDS handle control flow, such as iteration, from both the programming model and execution level perspectives. This survey will be of interest for both users and designers of DDS.

Fault-tolerance for distributed iterative dataflows in action

Distributed dataflow systems (DDS) are widely employed in graph processing and machine learning (ML), where many of these algorithms are iterative in nature. Typically, DDS achieve fault-tolerance using checkpointing mechanisms or they exploit algorithmic properties to enable fault-tolerance without the need for checkpoints. Recently, for graph processing, we proposed utilizing unblocking checkpointing , to parallelize the execution pipeline and checkpoint writing, as well as confined recovery , to enable fast recovery upon partial node failures. Furthermore, for ML algorithms implemented using broadcast variables, we proposed utilizing replica recovery , to leverage broadcast variable replicas and facilitate failure recovery checkpointing-free. In this demonstration, we showcase these fault-tolerance techniques using Apache Flink. Attendees will be able to: (i) run representative iterative algorithms including PageRank, Connected Components, and K-Means, (ii) explore the internal behavior of DDS under the influence of unblocking checkpointing, and (iii) trigger failures, to observe the effects of confined recovery and replica recovery.

On Fault Tolerance for Distributed Iterative Dataflow Processing

Large-scale graph and machine learning analytics widely employ distributed iterative processing. Typically, these analytics are a part of a comprehensive workflow, which includes data preparation, model building, and model evaluation. General-purpose distributed dataflow frameworks execute all steps of such workflows holistically. This holistic view enables these systems to reason about and automatically optimize the entire pipeline. Here, graph and machine learning analytics are known to incur a long runtime since they require multiple passes over the data until convergence is reached. Thus, fault tolerance and a fast-recovery from any intermittent failure is critical for efficient analysis. In this paper, we propose novel fault-tolerant mechanisms for graph and machine learning analytics that run on distributed dataflow systems. We seek to reduce checkpointing costs and shorten failure recovery times. For graph processing, rather than writing checkpoints that block downstream operators, our mechanism writes checkpoints in an unblocking manner that does not break pipelined tasks. In contrast to the conventional approach for unblocking checkpointing (e.g., that manage checkpoints independently for immutable datasets), we inject the checkpoints of mutable datasets into the iterative dataflow itself. Hence, our mechanism is iteration-aware by design. This simplifies the system architecture and facilitates coordinating checkpoint creation during iterative graph processing. Moreover, we are able to rapidly rebound, via confined recovery, by exploiting the fact that log files exist locally on healthy nodes and managing to avoid a complete recomputation from scratch. In addition, we propose replica recovery for machine learning algorithms, whereby we employ a broadcast variable that enables us to quickly recover without having to introduce any checkpoints. In order to evaluate our fault tolerance strategies, we conduct both a theoretical study and experimental analyses using Apache Flink and discover that they outperform blocking checkpointing and complete recovery.

Modeling and optimizing large-scale data flows

Modern scientific collaborations require large-scale integration of various processes. Higher-level dataflow languages are used on top of parallel and distributed dataflow systems to enable faster data-intensive workflow programs development, their easier optimization, and more maintainable code. In this paper, we present the rationales, design, and application of the needed advanced support for modeling and optimizing data flows for data mining and integration processes. The optimization research and development is based on dataflow pre-execution modeling and extending the registry of process activities by advanced annotations. Additionally, the overall process from a dynamic model to a static model as input for the optimization algorithms is described. This novel approach is implemented within an advanced graphical user interface, called the Process Designer, in order to support semi-automatic optimization as well as within a dataflow execution platform, called the Gateway. It can be adapted to any dataflow language implementation. The Process Designer architecture based on modern (meta-)modeling concepts naturally supports validated transformations between external textual and internal graphical representations of the targeted dataflow language, and in this way significantly increases the productivity and robustness of the implementation processes.

Two Tools for Interprocess Communication in Distributed Data-Flow Systems

Existing data-flow languages are incapable of supporting large, manageable and reliable non-deterministic programs. Operating system implementations and real-time applications require high-level non-deterministic programming facilities. Hoare's Communicating Sequential Processes and Brinch Hansen's Distributed Processes are two widely acclaimed distributed programming languages for conventional distributed computing systems. The powerful constructs of interprocess communication and synchronisation supported by the above two languages have been incorporated in Communicating Data Flow Channels (CDFC), and Distributed Data Flow Channels (DDFC), two languages proposed in this paper for distributed processing in data-flow systems. Several new features have been incorporated in these languages to make their execution more efficient in a data-flow environment. The implementation of these two languages on an abstract data-flow machine is investigated in detail. Two simulators have been implemented for testing the effectiveness of our implementation schemes for these two data-flow languages. These simulators are written in Pascal and run on a DEC 1090 computing system. This work demonstrates that a data-flow computing system can accommodate, without taking resort to specialised hardware units, the non-deterministic and synchronisation constructs of conventional languages with the attendant benefits of parallel data-driven execution.