To cope with the high resource (network and compute) demands of real-time video analytics pipelines, recent systems have relied on frame filtering. However, filtering has typically been done neural networks running edge/backend servers that are expensive to operate. This paper investigates on-camera filtering, which moves beginning pipeline. Unfortunately, we find commodity cameras limited compute resources only permit via differencing based low-level features. Used incorrectly, such...

A memory leak in a Java program occurs when object references that are no longer needed unnecessarily maintained. Such leaks difficult to understand because static analyses typically cannot precisely identify these redundant references, and existing dynamic for detection track report fine-grained information about individual objects, producing results usually hard interpret lack precision.

Continuous processing of a streaming graph maintains an approximate result the iterative computation on recent version graph. Upon user query, accurate current can be quickly computed by feeding results to --- form incremental that corrects (small amount of) error in result. Despite effectiveness this approach growing graphs, it is generally not applicable when edge deletions are present existing approximations lead either incorrect (e.g., monotonic computations terminate at minima/maxima)...

The past decade has witnessed the increasing demands on data-driven business intelligence that led to proliferation of data-intensive applications. A managed object-oriented programming language such as Java is often developer's choice for implementing applications, due its quick development cycle and rich community resource. While use languages makes easier, their automated memory management comes at a cost. When runtime meets Big Data, this cost significantly magnified becomes...

Many opportunities for easy, big-win, program optimizations are missed by compilers. This is especially true in highly layered Java applications. Often at the heart of these optimization lie computations that, with great expense, produce data values that have little impact on program's final output. Constructing a new date formatter to format every date, or populating large set full expensively constructed structures only check its size: involve costs out line benefits gained. disparity...

Many large-scale Java applications suffer from runtime bloat. They execute large volumes of methods, and create many temporary objects, all to relatively simple operations. There are opportunities for performance optimizations in these applications, but most being missed by existing optimization tooling technology. While JIT struggle a few percent, experts analyze deployed regularly find gains 2x or more.

Runtime bloat degrades significantly the performance and scalability of software systems. An important source is inefficient use containers. It expensive to create inefficiently-used containers invoke their associated methods, as this may ultimately execute large volumes code, with call stacks dozens deep, allocate many temporary objects.

Generally believed to be a problem belonging the compiler and architecture communities, performance optimization has rarely gained attention in mainstream software engineering research. However, due proliferation of large-scale object-oriented designed solve increasingly complex problems, issues stand out, preventing applications from meeting their requirements. Many such result design principles adopted widely research community, as idea reuse patterns. We argue that, modern era when...

Software tools for program understanding, transformation, verification, and testing often require an efficient yet highly-precise alias analysis. Typically this is done by computing points-to information, from which queries can be answered. This paper presents a novel context-sensitive, demand-driven analysis Java that achieves efficiency answering directly, instead of relying on underlying The formulated as context-free-language (CFL) reachability problem over language models calling...

Modern object-oriented software commonly suffers from runtime bloat that significantly affects its performance and scalability. Studies have shown one important pattern of is the work repeatedly done to compute same data values. Very often cost computation very high it thus beneficial memoize invariant values for later use. While this a common practice in real-world development, manually finding daunting task during development tuning. To help developers quickly find such optimization...

There is more than a decade-long history of using static analysis to find bugs in systems such as Linux. Most the existing analyses developed for these are simple checkers that based on pattern matching. Despite presence many sophisticated interprocedural analyses, few them have been employed improve code due their complex implementations and poor scalability. In this paper, we revisit scalability problem from "Big Data" perspective. That is, turn into Big Data analytics leverage novel data...

Graph mining is an important category of graph algorithms that aim to discover structural patterns such as cliques and motifs in a graph. While great deal work has been done recently on computation PageRank, systems support for scalable still limited. Existing Arabesque focus distributed computing need large amounts compute memory resources.We built RStream, the first single-machine, out-of-core system leverages disk store intermediate data. At its core are two innovations: (1) rich...

Most Big Data systems are written in managed languages, such as Java, C#, or Scala. These suffer from severe memory problems due to the massive volume of objects created process input data. Allocating and deallocating a sea data puts strain on existing garbage collectors (GC), leading high management overheads reduced performance.This paper describes design implementation Yak, friendly collector that provides throughput low latency for all JVM-based languages. Yak divides heap into control...

As aspect-oriented software development gains popularity, there is growing interest in using aspects to implement cross-cutting concerns object-oriented systems. When features are added an program, or when existing program modified, the new needs be regression tested validate these changes. To reduce cost of testing, a regression-test-selection technique can used select only necessary subset test cases rerun. Unfortunately, approaches for selection not effective presence aspectual...

A big source of run-time performance problems in large-scale, object-oriented applications is the frequent creation data structures (by same allocation site) whose lifetimes are disjoint, and shapes content always same. Constructing these computing values many times expensive; significant improvements can be achieved by reusing their instances, shapes, and/or rather than reconstructing them. This paper presents a technique that used to help programmers find sites create such improve...

In large programs written in managed languages such as Java and C#, holding unnecessary references often results memory leaks bloat, degrading significantly their run-time performance scalability. Despite the existence of many leak detectors for languages, these target low-level objects; a result, reports contain false warnings lack sufficient semantic information to help diagnose problems. This paper introduces specification-based technique called LeakChaser that can not only capture...

Over the past decade, increasing demands on data-driven business intelligence have led to proliferation of large-scale, data-intensive applications that often huge amounts data (often at terabyte or petabyte scale) process. An object-oriented programming language such as Java is developer's choice for implementing applications, primarily due its quick development cycle and rich community resource. While use languages makes easier, significant performance problems can be seen --- combination...

Real-world data-parallel programs commonly suffer from great memory pressure, especially when they are executed to process large datasets. Memory problems lead excessive GC effort and out-of-memory errors, significantly hurting system performance scalability. This paper proposes a systematic approach that can help tasks survive improving their scalability without needing any manual tune parameters. Our advocates interruptible task (ITask), new type of be interrupted upon pressure---with part...

Evolving graph processing involves repeating analyses, which are often iterative, over multiple snapshots of the corresponding to different points in time. Since an evolving share a great number vertices and edges, traditional approaches that process these one at time without exploiting this overlap contain much wasted effort on both data loading computation, making them extremely inefficient. In article, we identify major sources inefficiencies present two optimization techniques address...

Modern data-parallel systems such as Spark rely increasingly on in-memory computing that can significantly improve the efficiency of iterative algorithms. To process real-world datasets, modern often require extremely large amounts memory, which are both costly and energy-inefficient. Emerging non-volatile memory (NVM) technologies offers high capacity compared to DRAM low energy SSDs. Hence, NVMs have potential fundamentally change dichotomy between durable storage in Big Data processing....

In large programs written in managed languages such as Java and C#, holding unnecessary references often results memory leaks bloat, degrading significantly their run-time performance scalability. Despite the existence of many leak detectors for languages, these target low-level objects; a result, reports contain false warnings lack sufficient semantic information to help diagnose problems. This paper introduces specification-based technique called LeakChaser that can not only capture...