Efficient SIMD and MIMD parallelization of hash-based aggregation by conflict mitigation

Jiang, Peng; Agrawal, Gagan

doi:10.1145/3079079.3079080

Cited by 17 publications

(6 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For example, seismic [63], stencil [64], [65], electromagnetic [66], molecular dynamics [67], Fast Multipole Methods [68], tensors [39], deep learning [69], [70], databases [49], [71], [72], big data [73], systems and graph engines [74], and many more.…”

Section: State-of-the-art Shared-memory Optimizationsmentioning

confidence: 99%

“…Similarly, MapReduce framework is optimized for KNC architecture in [73], in which the thread-level parallelism is leveraged. Vectorization through explicit SIMDization is explored in many database primitive operations including hashing [71], [72], as well as sequential scans, aggregation, index operations, and joins [49].…”

Section: State-of-the-art Shared-memory Optimizationsmentioning

confidence: 99%

“…Some of this work is inherited and customized to our application code. For instance, SoA of [68], AoSoA of [22], low-level, MCDRAM-aware allocator of [39], data dependency conflicts migration of [71], Hilbert-based recursive tiling/blocking of [74], cache line aware optimization of [1], [46], [75], and partial coloring of [79]. In our work, we deal with irregular memory access patterns through optimizing for the cache line awareness based upon minimizing memory reference arithmetic and pointer chasing, as well as localizing a large bulk of computations inside a compute core.…”

Section: State-of-the-art Shared-memory Optimizationsmentioning

confidence: 99%

See 2 more Smart Citations

Optimizations of Unstructured Aerodynamics Computations for Many-core Architectures

Farhan

Keyes

2018

IEEE Trans. Parallel Distrib. Syst.

View full text Add to dashboard Cite

Section: State-of-the-art Shared-memory Optimizationsmentioning

confidence: 99%

Section: State-of-the-art Shared-memory Optimizationsmentioning

confidence: 99%

Section: State-of-the-art Shared-memory Optimizationsmentioning

confidence: 99%

See 1 more Smart Citation

Optimizations of Unstructured Aerodynamics Computations for Many-core Architectures

Farhan

Keyes

2018

IEEE Trans. Parallel Distrib. Syst.

View full text Add to dashboard Cite

“…Data processing and database algorithms fit particularly well into this pattern. Their behavior is often highly predictable; and there are many examples that show how the awareness of resources and their uses in database code can significantly improve performance [1,14,15,18,26,31,34]. Fig.…”

Section: Task Model: Mxtasksmentioning

confidence: 99%

mxkernel: A Novel System Software Stack for Data Processing on Modern Hardware

et al. 2020

View full text Add to dashboard Cite

Emerging hardware platforms are characterized by large degrees of parallelism, complex memory hierarchies, and increasing hardware heterogeneity. Their theoretical peak data processing performance can only be unleashed if the different pieces of systems software collaborate much more closely and if their traditional dependencies and interfaces are redesigned. We have developed the key concepts and a prototype implementation of a novel system software stack named mxkernel. For MxKernel, efficient large scale data processing capabilities are a primary design goal. To achieve this, heterogeneity and parallelism become first-class citizens and deep memory hierarchies are considered from the very beginning. Instead of a classical “thread” model, mxkernel provides a simpler control flow abstraction: mxtasks model closed units of work, for which mxkernel will guarantee the required execution semantics, such exclusive access to a specific object in memory. They can be a very elegant abstraction also for heterogeneity and resource sharing. Furthermore, mxtasks are annotated with metadata, such as code variants (to support heterogeneity), memory access behavior (to improve cache efficiency and support memory hierarchies), or dependencies between mxtasks (to improve scheduling and avoid synchronization cost). With precisely the required metadata available, mxkernel can provide a lightweight, yet highly efficient form of resource management, even across applications, operating system, and database. Based on the mxkernel prototype we present preliminary results from this ambitious undertaking. We argue that threads are an ill-suited control flow abstraction for our modern computer architectures and that a task-based execution model is to be favored.

show abstract

“…The first step is an optional local aggregation where data is aggregated locally, followed by a second step where data is repartitioned and transferred to the final destination node for aggregation [45,14]. The local aggregation can reduce the amount of data transferred in the second step for algebraic aggregations, as tuples with the same GROUP BY key are aggregated to a single tuple during local aggregation [6,52,22,35,48]. Local aggregation works effectively for low-cardinality domains, such as age, sex or country, where data can be reduced substantially and make the cost of the repartition step negligible.…”

Section: Introductionmentioning

confidence: 99%

Chasing Similarity: Distribution-aware Aggregation Scheduling (Extended Version)

Liu¹,

Salmasi²,

Blanas³

et al. 2018

Preprint

View full text Add to dashboard Cite

Parallel aggregation is a ubiquitous operation in data analytics that is expressed as GROUP BY in SQL, reduce in Hadoop, or segment in TensorFlow. Parallel aggregation starts with an optional local pre-aggregation step and then repartitions the intermediate result across the network. While local preaggregation works well for low-cardinality aggregations, the network communication cost remains significant for highcardinality aggregations even after local pre-aggregation. The problem is that the repartition-based algorithm for highcardinality aggregation does not fully utilize the network.In this work, we first formulate a mathematical model that captures the performance of parallel aggregation. We prove that finding optimal aggregation plans from a known data distribution is NP-hard, assuming the Small Set Expansion conjecture. We propose GRASP, a GReedy Aggregation Scheduling Protocol that decomposes parallel aggregation into phases. GRASP is distribution-aware as it aggregates the most similar partitions in each phase to reduce the transmitted data size in subsequent phases. In addition, GRASP takes the available network bandwidth into account when scheduling aggregations in each phase to maximize network utilization. The experimental evaluation on real data shows that GRASP outperforms repartition-based aggregation by 3.5× and LOOM by 2.0×.

show abstract

Efficient SIMD and MIMD parallelization of hash-based aggregation by conflict mitigation

Cited by 17 publications

References 22 publications

Optimizations of Unstructured Aerodynamics Computations for Many-core Architectures

Optimizations of Unstructured Aerodynamics Computations for Many-core Architectures

mxkernel: A Novel System Software Stack for Data Processing on Modern Hardware

Chasing Similarity: Distribution-aware Aggregation Scheduling (Extended Version)

Contact Info

Product

Resources

About