Co-location, which deploys long running applications and batch-processing applications in the same computing cluster, has become a promising way to improve resource utility for large cloud datacenters. However, co-loc...
详细信息
Co-location, which deploys long running applications and batch-processing applications in the same computing cluster, has become a promising way to improve resource utility for large cloud datacenters. However, co-location brings huge challenges to task scheduling because different types of workloads may affect each other. Existing works on task scheduling rarely focus on the scenario of co-location. This article presents Co-ScheRRL, a scheduling algorithm delicately designed for co-located workloads. Co-ScheRRL consists of two major mechanisms: i) a self-attention encoding mechanism which encodes and represents states of the computing cluster as a set of embedding feature vectors;ii) a deep reinforcement learning (DRL) relational reasoning mechanism which calculates and compares different scheduling actions under different co-located workloads pattern via DRL feedback reward signals based on these feature vectors. Our two mechanisms can tackle complicatedly and dynamically varying behaviors of co-located workloads. With the help of these two mechanisms, Co-ScheRRL is able to construct high-quality scheduling policies. Trace-driven simulation demonstrates that Co-ScheRRL outperforms existing scheduling algorithms in terms of makespan by more than 38.4% and throughput by more than 166.7%.
In systems of multiple autonomous underwater vehicles (AUVs), to achieve cooperative operation and cluster intelligence, information is often disseminated via broadcasting. However, due to the long propagation delay a...
详细信息
In systems of multiple autonomous underwater vehicles (AUVs), to achieve cooperative operation and cluster intelligence, information is often disseminated via broadcasting. However, due to the long propagation delay and slow transmission rate of underwater acoustic communication, traditional broadcast scheduling algorithms require a long broadcast period to avoid signal collision. To improve the channel utilization rate as much as possible and improve the update rate for broadcast information, we propose an optimal broadcast scheduling algorithm. This algorithm uses the location information of AUVs to adjust the broadcast sequence and broadcast schedule, to achieve the shortest possible collision-free broadcast period in the broadcast network for the current node distribution. Simulation experiments show that this algorithm can achieve a broadcast period much shorter than that of traditional TDMA and higher channel utilization without signal collision. In addition, the simulations prove the feasibility of applying the algorithm in an actual MAC protocol.
Efficient bus scheduling is a crucial component for the improvement of public transit services. Without systematic optimization and efficient shift arrangement, the bus scheduling system may suffer from poor vehicle l...
详细信息
Efficient bus scheduling is a crucial component for the improvement of public transit services. Without systematic optimization and efficient shift arrangement, the bus scheduling system may suffer from poor vehicle loading rate or crowd onboard, resulting in wasted energy and passenger dissatisfaction. In this paper, we consider a crowdsourced bus service system (on a fixed route) that receives user requests as input and computes the scheduling of buses with flexible departure time and skip-stop to minimize the travel time of users. We first show that the general problem of computing the optimal scheduling is NP-hard. On the other hand, for the case when skip-stop is not adopted, we propose the Optimized Departure Time (ODT) algorithm that computes optimal scheduling. Our algorithm is built on an innovative reduction of the problem to some variants of the k-clustering problem and an efficient application of dynamic programming. On top of ODT, we further improve the effectiveness of the solution by utilizing the power of skip-stop tactic, named ODTS. Our experimental results demonstrate that ODT and ODTS dramatically outperform existing algorithms for the bus scheduling problem in terms of effectiveness and efficiency. Moreover, the solutions given by ODTS are very close to the optimum.
To handle task execution, modern supercomputers employ thousands (or millions) of processors. In such supercomputers, task scheduling has a meaningful impression on system performance. To improve efficiency, task sche...
详细信息
To handle task execution, modern supercomputers employ thousands (or millions) of processors. In such supercomputers, task scheduling has a meaningful impression on system performance. To improve efficiency, task scheduling algorithms aim to decrease the volume of communication and the number of message exchanges. These efforts, however, result in other bottlenecks, such as high-link congestion. In addition, the heterogeneity of processors and networks is another major challenge for schedulers. This paper presents a new algorithm for scheduling called Heterogeneity-Aware Task scheduling (HATS). The proposed algorithm adopts an updated multi-level hyper-graph partitioning approach. It describes a new method of aggregation in the coarsening step that helps to accurately coarsen the hyper-graph of the task model. The Raccoon Optimization algorithm is then used in the initial partitioning phase, and in the un-coarsening phase, a novel refinement procedure optimises the initial partitions. The experiments on this approach showed that, compared to the other well-known algorithms, the proposed method offers better schedules with lower communication volume and imbalance ratio in a shorter time.
Time-Sensitive Networking (TSN) can ensure deterministic communications for time-critical traffic, which plays a crucial role in various real-time scenarios. In this letter, we propose and study a neglected problem in...
详细信息
Time-Sensitive Networking (TSN) can ensure deterministic communications for time-critical traffic, which plays a crucial role in various real-time scenarios. In this letter, we propose and study a neglected problem in TSN, named flow ordering problem, which provides a new perspective on improving the scheduling of large-scale TSN. Specifically, we formulate the flow ordering problem, look into its theoretical basis, and prove this problem is NP-hard. Furthermore, we propose a hybrid search algorithm to provide an optimized scheduling order. Simulation results verify the significant impact of the flow ordering problem on TSN scheduling and the effectiveness of our algorithm.
Establishing an efficient cloud computing task scheduling model is the object of many scholars' research. In view of the low scheduling efficiency in cloud computing task scheduling, we propose a cloud computing t...
详细信息
The scale of data centres and network traffic at the core of modern information service infrastructure is increasing. At present, there are many problems in network traffic management under the new architecture, such ...
详细信息
scheduling packets with end-to-end deadline constraints in multihop networks is an important problem that has been notoriously difficult to tackle. Recently, there has been progress on this problem in the worst-case t...
详细信息
scheduling packets with end-to-end deadline constraints in multihop networks is an important problem that has been notoriously difficult to tackle. Recently, there has been progress on this problem in the worst-case traffic setting, with the objective of maximizing the number of packets delivered within their deadlines. Specifically, the proposed algorithms were shown to achieve Omega(1/log(L)) fraction of the optimal objective value if the minimum link capacity in the network is C-min = Omega( log(L)), where L is the maximum length of a packet's route in the network (which is bounded by the packet's maximum deadline). However, such guarantees can be quite pessimistic due to the strict worst-case traffic assumption and may not accurately reflect real-world settings. In this work, we aim to address this limitation by exploring whether it is possible to design algorithms that achieve a constant fraction of the optimal value while relaxing the worst-case traffic assumption. We provide a positive answer by demonstrating that in stochastic traffic settings, such as i.i.d. packet arrivals, near-optimal, (1 - epsilon)-approximation algorithms can be designed if C-min = Omega (log(L/epsilon)epsilon(2)). To the best of our knowledge, this is the first result that shows this problem can be solved near-optimally under nontrivial assumptions on traffic and link capacity. We further present extended simulations using real network traces with non-stationary traffic, which demonstrate that our algorithms outperform worst-case-based algorithms in practical settings.
This paper investigates a dynamic packet scheduling algorithm designed to enhance the eXtended Reality (XR) capacity of fifth-generation (5G)-Advanced networks with multiple cells, users, and services. The scheduler e...
详细信息
ISBN:
(纸本)9781728190549
This paper investigates a dynamic packet scheduling algorithm designed to enhance the eXtended Reality (XR) capacity of fifth-generation (5G)-Advanced networks with multiple cells, users, and services. The scheduler exploits the newly defined protocol data unit (PDU)-set information for XR traffic flows to enhance its quality-of-service awareness. To evaluate the performance of the proposed solution, advanced dynamic system-level simulations are conducted. The findings reveal that the proposed scheduler offers a notable improvement in increasing XR capacity up to 45%, while keeping the same enhanced mobile broadband (eMBB) cell throughput as compared to the well-known baseline schedulers.
In the planning phase of collaborative development projects, a reasonable scheduling plan plays a pivotal role in shortening development cycle and reducing costs. However, in the traditional collaborative development ...
详细信息
暂无评论