An increasingly important system requirement for distributed stream processing applications is to provide strong correctness guarantees under unexpected failures and out-of-order data so that its results can be authoritative (not needing complementary batch results). Although existing systems have put a lot of effort into addressing some specific issues, such as consistency and completeness, how to enable users to make flexible and transparent trade-off decisions among correctness, performance, and cost still remains a practical challenge. Specifically, similar mechanisms are usually applied to tackle both consistency and completeness, which can result in unnecessary performance penalties. We present Apache Kafka's core design for stream processing, which relies on its persistent log architecture as the storage and inter-processor communication layers to achieve correctness guarantees. Kafka Streams, a scalable stream processing client library in Apache Kafka, defines the processing logic as read-process-write cycles in which all processing state updates and result outputs are captured as log appends. Idempotent and transactional write protocols are utilized to guarantee exactly-once semantics. Furthermore, revision-based speculative processing is employed to emit results as soon as possible while handling out-of-order data. We also demonstrate how Kafka Streams behaves in practice with large-scale deployments and performance insights exhibiting its flexible and low-overhead trade-offs.

Aeolus is a prototype implementation of a topology optimizer on top of the distributed streaming system Storm. Aeolus extends Storm with a batching layer which can increase the topology's throughput by more than one order of magnitude. Furthermore, Aeolus implements an optimization algorithm that computes the optimal batch size and degree of parallelism for each node in the topology automatically. Even if Aeolus is built on top of Storm, the developed concepts are not limited to Storm and can be applied to any distributed intra-node-parallel streaming system. We propose to demo Aeolus using an interactive Web UI. One part of the Web UI is a topology builder allowing the user to interact with the system. Topologies can be created from scratch and their structure and/or parameters can be modified. Furthermore, the user is able to observe the impact of the changes on the optimization decisions and runtime behavior. Additionally, the Web UI gives a deep insight in the optimization process by visualizing it. The user can interactively step through the optimization process while the UI shows the optimizer's state, computations, and decisions. The Web UI is also able to monitor the execution of a non-optimized and optimized topology simultaneously showing the advantage of using Aeolus.

Many systems for big data analytics employ a data flow abstraction to define parallel data processing tasks. In this setting, custom operations expressed as user-defined functions are very common. We address the problem of performing data flow optimization at this level of abstraction, where the semantics of operators are not known. Traditionally, query optimization is applied to queries with known algebraic semantics. In this work, we find that a handful of properties, rather than a full algebraic specification, suffice to establish reordering conditions for data processing operators. We show that these properties can be accurately estimated for black box operators by statically analyzing the general-purpose code of their user-defined functions. We design and implement an optimizer for parallel data flows that does not assume knowledge of semantics or algebraic properties of operators. Our evaluation confirms that the optimizer can apply common rewritings such as selection reordering, bushy join-order enumeration, and limited forms of aggregation push-down, hence yielding similar rewriting power as modern relational DBMS optimizers. Moreover, it can optimize the operator order of nonrelational data flows, a unique feature among today's systems.