Accelerated Work Stealing

Larkins, D. Brian; Snyder, John; Dinan, James

doi:10.1145/3337821.3337878

Cited by 5 publications

(3 citation statements)

References 37 publications

(43 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The proposed solution uses work stealing [9,10] to mitigate workload on congested servers to the benefit of other servers. Work stealing involves an idle processor stealing one or more tasks from its neighbor's waiting list [9,11,12]. This method already proved its ability to achieve lower latencies in single parallel job execution on multiple processors [9,10,13].…”

Section: Data Planementioning

confidence: 99%

WoS-CoMS: Work Stealing-Based Congestion Management Scheme for SDN Programmable Networks

Yankam,

Tchendji,

Myoupo

2024

J Netw Syst Manage

View full text Add to dashboard Cite

In recent years, the SDN (Software-Defined Networking) paradigm emerged as an easy way to manage large-scale network infrastructures through programmability brought out and its control plane/data plane decoupling logic. This enables infrastructure and service providers to have a global view of the network and track traffic flows from a remote controller. However, congestion control remains a concern due to the evolution of increasingly complex and resource-intensive user requirements (virtual reality, metaverse, Internet of Things (IoT), Artificial Intelligence (AI), Cloud, ...) on network infrastructures. This server state leads to high latency in request processing and data loss. This paper proposes in such controller-supervised environment, a congestion management scheme within network service servers to maintain acceptable quality of service. The strategy relies on work stealing to ensure better workload balancing. Simulations show that the proposed solution can reduce congestion load into the servers by up to 22%, depending on request grain size, within a shorter latency than other works in the literature. Moreover, the proposed solution allows stolen tasks to be completed within a shorter timeframe.

show abstract

Section: Data Planementioning

confidence: 99%

WoS-CoMS: Work Stealing-Based Congestion Management Scheme for SDN Programmable Networks

Yankam,

Tchendji,

Myoupo

2024

J Netw Syst Manage

View full text Add to dashboard Cite

show abstract

“…33 Larkins et al have introduced an alternative of one-sided RDMA communication called Portals interface 34 to accelerate work stealing in distributed memory. 35 Regarding the approach for reducing migration overhead, Lifflander et al introduced a hierarchical technique that applies the persistence principle to distribute the load of task-based applications. 36 Menon et al proposed using partial information about the global system state to improve stealing decisions as well as balance the load by randomized work-stealing.…”

Section: Related Workmentioning

confidence: 99%

“…Dinan et al have designed a scalable model for work‐stealing using PGAS by the Aggregate Remote Memory Copy Interface (ARMCI), 32 focusing on techniques to reduce locking on the critical path and contention of splitting work 33 . Larkins et al have introduced an alternative of one‐sided RDMA communication called Portals interface 34 to accelerate work stealing in distributed memory 35 . Regarding the approach for reducing migration overhead, Lifflander et al introduced a hierarchical technique that applies the persistence principle to distribute the load of task‐based applications 36 .…”

Section: Related Workmentioning

confidence: 99%

From reactive to proactive load balancing for task‐based parallel applications in distributed memory machines

Chung

Weidendorfer

Fürlinger

et al. 2023

Concurrency and Computation

View full text Add to dashboard Cite

SummaryLoad balancing is often a challenge in task‐parallel applications. The balancing problems are divided into static and dynamic. “Static” means that we have some prior knowledge about load information and perform balancing before execution, while “dynamic” must rely on partial information of the execution status to balance the load at runtime. Conventionally, work stealing is a practical approach used in almost all shared memory systems. In distributed memory systems, the communication overhead can make stealing tasks too late. To improve, people have proposed a reactive approach to relax communication in balancing load. The approach leaves one dedicated thread per process to monitor the queue status and offload tasks reactively from a slow to a fast process. However, reactive decisions might be mistaken in high imbalance cases. First, this article proposes a performance model to analyze reactive balancing behaviors and understand the bound leading to incorrect decisions. Second, we introduce a proactive approach to improve further balancing tasks at runtime. The approach exploits task‐based programming models with a dedicated thread as well, namely . Nevertheless, the main idea is to force not only to monitor load; it will characterize tasks and train load prediction models by online learning. “Proactive” indicates offloading tasks before each execution phase proactively with an appropriate number of tasks at once to a potential victim (denoted by an underloaded/fast process). The experimental results confirm speedup improvements from to in important use cases compared to the previous solutions. Furthermore, this approach can support co‐scheduling tasks across multiple applications.

show abstract

Proactive Task Offloading for Load Balancing in Iterative Applications

Chung¹,

Weidendorfer²,

Fürlinger³

et al. 2023

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Load imbalance is often a challenge for applications in parallel systems. Static cost models and pre-partitioning algorithms distribute the load at the beginning. Nevertheless, dynamic changes during execution or inaccurate cost indicators may lead to imbalance at runtime. Reactive work-stealing strategies can help monitor the execution and perform task migration to balance the load. However, the benefits depend on migration overhead and assumption about future execution.Our proactive approach further improves existing solutions by applying machine learning to online load prediction. Following that, we propose a fully distributed algorithm for adapting the prediction result to guide task offloading. The experiments are performed with an artificial test case and a realistic application named Sam(oa)$$^2$$ 2 on three systems with different communication overhead. Our results confirm improvements for important use cases compared to previous solutions. Furthermore, this approach can support co-scheduling tasks across multiple applications.

show abstract

Accelerated Work Stealing

Cited by 5 publications

References 37 publications

WoS-CoMS: Work Stealing-Based Congestion Management Scheme for SDN Programmable Networks

WoS-CoMS: Work Stealing-Based Congestion Management Scheme for SDN Programmable Networks

From reactive to proactive load balancing for task‐based parallel applications in distributed memory machines

Proactive Task Offloading for Load Balancing in Iterative Applications

Contact Info

Product

Resources

About