Optimal latency-throughput tradeoffs for data parallel pipelines

Subhlok, Jaspar; Vondran, Gary

doi:10.1145/237502.237508

Cited by 85 publications

(69 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…If the computations of a given stage are independent from one data set to another, two consecutive computations (different data sets) for the same stage can be mapped onto distinct processors, thus reducing the period for the processing of this stage. Such a stage can be replicated, using the terminology of Subhlok and Vondran [27,28] and of the DataCutter team [6,7,26]. This corresponds to the dealable stages of Cole [11].…”

Section: Working Out An Examplementioning

confidence: 99%

See 1 more Smart Citation

Complexity Results for Throughput and Latency Optimization of Replicated and Data-parallel Workflows

Benoît

Robert

2008

Algorithmica

View full text Add to dashboard Cite

Mapping applications onto parallel platforms is a challenging problem, even for simple application patterns such as pipeline or fork graphs. Several antagonist criteria should be optimized for workflow applications, such as throughput and latency (or a combination). In this paper, we consider a simplified model with no communication cost, and we provide an exhaustive list of complexity results for different problem instances. Pipeline or fork stages can be replicated in order to increase the throughput by sending consecutive data sets onto different processors. In some cases, stages can also be data-parallelized, i.e. the computation of one single data set is shared between several processors. This leads to a decrease of the latency and an increase of the throughput. Some instances of this simple model are shown to be NP-hard, thereby exposing the inherent complexity of the mapping problem. We provide polynomial algorithms for other problem instances. Altogether, we provide solid theoretical foundations for the study of mono-criterion or bi-criteria mapping optimization problems.

show abstract

Section: Working Out An Examplementioning

confidence: 99%

“…For example Subhlok and Vondran [27,28] have been able to design dynamic pro-gramming algorithms for bi-criteria mappings on homogeneous platforms. Also, if we neglect all communication costs, minimizing the period amounts to solve the well-known chains-to-chains problem.…”

Section: Introductionmentioning

confidence: 99%

Complexity Results for Throughput and Latency Optimization of Replicated and Data-parallel Workflows

Benoît

Robert

2008

Algorithmica

View full text Add to dashboard Cite

show abstract

“…Subhlok and Vondran [10] have proposed a dynamic programming solution for optimizing latency under throughput constraints for applications composed of a chain of data-parallel tasks. Benoit and Robert [11] study the theoretical complexity of latency and throughput optimization of pipeline and fork graphs with replication and data-parallelism under the assumptions of linear clustering and round-robin processing of input data items.…”

Section: Related Workmentioning

confidence: 99%

Toward Optimizing Latency Under Throughput Constraints for Application Workflows on Clusters

Vydyanathan¹,

Çatalyürek

Kurç

et al. 2007

Euro-Par 2007 Parallel Processing

View full text Add to dashboard Cite

Abstract. In many application domains, it is desirable to meet some user-defined performance requirement while minimizing resource usage and optimizing additional performance parameters. For example, application workflows with realtime constraints may have strict throughput requirements and desire a low latency or response-time. The structure of these workflows can be represented as directed acyclic graphs of coarse-grained application tasks with data dependences. In this paper, we develop a novel mapping and scheduling algorithm that minimizes the latency of workflows that act on a stream of input data, while satisfying throughput requirements. The algorithm employs pipelined parallelism and intelligent clustering and replication of tasks to meet throughput requirements. Latency is minimized by exploiting task parallelism and reducing communication overheads. Evaluation using synthetic benchmarks and application task graphs shows that our algorithm 1) consistently meets throughput requirements even when other existing schemes fail, 2) produces lower-latency schedules, and 3) results in lesser resource usage.

show abstract

“…There are many prior works that use similar models to estimate throughput for linear chains of tasks [13] [14].…”

Section: Illustrating Examplesmentioning

confidence: 99%

“…There is a significant body of prior work in partitioning and mapping for high throughput [13] [20] have also been exploring solutions of partitioning applications for network processors. In contrast to the parallel program that we use in our system, the IXP-C compiler assumes that users develop large sequential programs.…”

Section: Related Workmentioning

confidence: 99%

A Throughput-Driven Task Creation and Mapping for Network Processors

Liu¹,

Li²,

Chen

2007

High Performance Embedded Architectures and Compilers

View full text Add to dashboard Cite

Abstract. Network processors are programmable devices that can process packets at a high speed. A network processor is typified by multithreading and heterogeneous multiprocessing, which usually requires programmers to manually create multiple tasks and map these tasks onto different processing elements. This paper addresses the problem of automating task creation and mapping of network applications onto the underlying hardware to maximize their throughput. We propose a throughput cost model to guide the task creation and mapping with the objective of both minimizing the number of stages in the processing pipeline and maximizing the average throughput of the slowest task simultaneously. The average throughput is modeled by taking communication cost, computation cost, memory access latency and synchronization cost into account. We envision that programmers write small functions for network applications, such that we use grouping and duplication to construct tasks from the functions. The optimal solution of creating tasks from m functions and mapping them to n processors is an NP-hard problem. Therefore, we present a practical and efficient heuristic algorithm with an O((n + m)m) complexity and show that the obtained solutions produce excellent performance for typical network applications. The entire framework has been implemented in the Open Research Compiler (ORC) adapted to compile network applications written in a domain-specific dataflow language. Experimental results show that the code produced by our compiler can achieve the 100% throughput on the OC-48 input line rate. OC-48 is a fiber optic connection that can handle a 2.488Gbps connection speeds, which is what our targeted hardware was designed for. We also demonstrate the importance of good creation and mapping choices on achieving high throughput. Furthermore, we show that reducing communication cost and efficient resource management are the most important factors for maximizing throughput on the Intel IXP network processors.

show abstract

Optimal latency-throughput tradeoffs for data parallel pipelines

Abstract: This paper addressesoptimal mapping of parallel programs composed of a chain of data parallel tasks onto the proces-

Cited by 85 publications

References 14 publications

Complexity Results for Throughput and Latency Optimization of Replicated and Data-parallel Workflows

Complexity Results for Throughput and Latency Optimization of Replicated and Data-parallel Workflows

Toward Optimizing Latency Under Throughput Constraints for Application Workflows on Clusters

A Throughput-Driven Task Creation and Mapping for Network Processors

Contact Info

Product

Resources

About