Stream Processing Engines (SPEs) are used for realtime and continuous processing with stateful operations. This type of processing poses numerous challenges due to its associated complexity, unpredictable input, and need for timely results. As a result, users tend to overprovision resources, and online scaling is required in order to overcome overloaded situations. Current attempts for expediting stateful processing are impractical, due to their inability to uphold the quality of results, maintain performance, and reduce memory requirements. In this paper, we present the SPEAr system, which can expedite processing of stateful operations automatically by trading accuracy for performance. SPEAr detects when it can accelerate processing by employing online sampling and accuracy estimation at no additional cost. We built SPEAr on top of Storm and our experiments indicate that it can reduce processing times by more than an order of magnitude, use more than an order of magnitude less memory, and offer accuracy guarantees in real-world benchmarks.

Load shedding is a technique that aims to ameliorate the consequences of the Velocity and the Volume of Big Data stream processing. When temporal input spikes appear, tuples are shed until a Stream Processing Engine's (SPE) processing capacity is not overwhelmed and results are produced in a timely fashion. Existing load shedding techniques have become obsolete and are not applicable to modern use-cases which require the extraction of patterns from continuously evolving (i.e., Variable) voluminous streams.In this work, we identify the shortcomings of existing load shedding techniques when applied to streams with concept drift. We propose Concept-Driven load shedding (CoD), which aims at limiting the data volume imposed on the SPE while producing high accuracy results. On top of that, we designed CoD for modern SPEs and made its overhead negligible. Our experiments indicate that CoD can deliver more than 10x more accurate results compared to the state of the art in load shedding. Also, CoD can offer up to 2.25× better performance compared to normal processing and reduce the processed data volume significantly.

Stream processing has become the dominant processing model for monitoring and real-time analytics. Modern Parallel Stream Processing Engines (pSPEs) have made it feasible to increase the performance in both monitoring and analytical queries by parallelizing a query's execution and distributing the load on multiple workers. A determining factor for the performance of a pSPE is the partitioning algorithm used to disseminate tuples to workers. Until now, partitioning methods in pSPEs have been similar to the ones used in parallel databases and only recently load-aware algorithms have been employed to improve the effectiveness of parallel execution. We identify and demonstrate the need to incorporate aggregation costs in the partitioning model when executing stateful operations in parallel, in order to minimize the overall latency and/or through-put. Towards this, we propose new stream partitioning algorithms, that consider both tuple imbalance and aggregation cost. We evaluate our proposed algorithms and show that they can achieve up to an order of magnitude better performance, compared to the current state of the art.

The growing popularity of the JSON format has fueled increased interest in loading and processing JSON data within analytical data processing systems. However, in many applications, JSON parsing dominates performance and cost. In this paper, we present a new JSON parser called Mison that is particularly tailored to this class of applications, by pushing down both projection and filter operators of analytical queries into the parser. To achieve these features, we propose to deviate from the traditional approach of building parsers using finite state machines (FSMs). Instead, we follow a two-level approach that enables the parser to jump directly to the correct position of a queried field without having to perform expensive tokenizing steps to find the field. At the upper level, Mison speculatively predicts the logical locations of queried fields based on previously seen patterns in a dataset. At the lower level, Mison builds structural indices on JSON data to map logical locations to physical locations. Unlike all existing FSM-based parsers, building structural indices converts control flow into data flow, thereby largely eliminating inherently unpredictable branches in the program and exploiting the parallelism available in modern processors. We experimentally evaluate Mison using representative real-world JSON datasets and the TPC-H benchmark, and show that Mison produces significant performance benefits over the best existing JSON parsers; in some cases, the performance improvement is over one order of magnitude.

There is a need to integrate SQL processing with more advanced machine learning (ML) analytics to drive actionable insights from large volumes of data. As a first step towards this integration, we study how to efficiently connect big SQL systems (either MPP databases or new-generation SQL-on-Hadoop systems) with distributed big ML systems. We identify two important challenges to address in the integrated data analytics pipeline: data transformation, how to efficiently transform SQL data into a form suitable for ML, and data transfer, how to efficiently handover SQL data to ML systems. For the data transformation problem, we propose an In-SQL approach to incorporate common data transformations for ML inside SQL systems through extended user-defined functions (UDFs), by exploiting the massive parallelism of the big SQL systems. We propose and study a general method for transferring data between big SQL and big ML systems in a parallel streaming fashion. Furthermore, we explore caching intermediate or final results of data transformation to improve the performance. Our techniques are generic: they apply to any big SQL system that supports UDFs and any big ML system that uses Hadoop InputFormats to ingest input data.

Data Stream Management Systems (DSMS) are crucial for modern high-volume/high-velocity data-driven applications, necessitating a distributed approach to processing them. In addition, data providers often require certain levels of confidentiality for their data, especially in cases of user-generated data, such as those coming out of physical activity/health tracking devices (i.e., our motivating application). This demonstration will showcase Synefo, an infrastructure that enables elastic scaling of DSMS operators, and CryptStream, a framework that provides confidentiality and access controls for data streams while allowing computation on untrusted servers, fused as CE-Storm. We will demonstrate both systems working in tandem and also visualize their behavior over time under different scenarios.

The slow intelligence system (SIS) technology is a novel technology for the design of a complex information system that is aware of the environment through multiple sensors and capable of improving its performance over time. In this paper we describe a practical slow intelligence system test bed where super-components can be specified to describe interactions among components. These super-components are automatically transformed into time controllers for components so they can be managed by the SIS test bed. We illustrate the application of this methodology to personal healthcare system design.

Live migration is a versatile option when it comes to attain load-balancing in IaaS-cloud architectures. Liveness, reliability and conformance to SLAs may all be achieved by moving a VM that creates excessive work from its current physical machine (PM) to a less busy node. Despite its promising features, live migration is an expensive operation in terms of resources. The situation gets further exacerbated when the movement involves PMs working off different file-systems which is often the case in shared-nothing IaaS-cloud infrastructures. In this paper, we suggest an approach that adapts the migration operation based on the I/O activity of the originating-VM. We introduce LiveFS, a FUSE-file system which traps all I/Os and helps determine how to best ship virtual disk segments across PMs in a share-nothing IaaS-cloud. Through prototyping and experimentation, we show that LiveFS can improve the shipment of VMs for diverse types of workloads. In particular, LiveFS succeeds in reducing the Total Migration Time by up to 64% compared to the "pre-copy" live migration technique. Furthermore during migration, we attain up-to 19% less I/O-delay if compared to the "post-copy" live-migration approach.