相关论文: Lossy Bulk Synchronous Parallel Processing Model f…

Stochastic Gradient Descent (SGD) has become the de facto way to train deep neural networks in distributed clusters. A critical factor in determining the training throughput and model accuracy is the choice of the parameter synchronization…

The main goal of distribution network (DN) expansion planning is essentially to achieve minimal investment constrained with specified reliability requirements. The reliability-constrained distribution network planning (RcDNP) problem can be…

According to the increasing complexity of network application and internet traffic, network processor as a subset of embedded processors have to process more computation intensive tasks. By scaling down the feature size and emersion of chip…

Deep Neural Networks (DNNs) have revolutionized numerous applications, but the demand for ever more performance remains unabated. Scaling DNN computations to larger clusters is generally done by distributing tasks in batch mode using…

This paper provides an in-depth characterization of GPU-accelerated systems, to understand the interplay between overlapping computation and communication which is commonly employed in distributed training settings. Due to the large size of…

The increasing use of middleboxes (e.g., NATs, firewalls) in the Internet has made it harder and harder to deploy new transport or higher layer protocols, or even extensions to existing ones. Current work to address this Internet transport…

Emerging 5G systems will need to efficiently support both enhanced mobile broadband traffic (eMBB) and ultra-low-latency communications (URLLC) traffic. In these systems, time is divided into slots which are further sub-divided into…

Reducing the average memory access time is crucial for improving the performance of applications running on multi-core architectures. With workload consolidation this becomes increasingly challenging due to shared resource contention.…

Planning under uncertainty is critical for robust robot performance in uncertain, dynamic environments, but it incurs high computational cost. State-of-the-art online search algorithms, such as DESPOT, have vastly improved the computational…

Large-scale deep learning models contribute to significant performance improvements on varieties of downstream tasks. Current data and model parallelism approaches utilize model replication and partition techniques to support the…

An important ingredient of the future 5G systems will be Ultra-Reliable Low-Latency Communication (URLLC). A way to offer URLLC without intervention in the baseband/PHY layer design is to use interface diversity and integrate multiple…

Minimizing the number of dropped User Datagram Protocol (UDP) messages in a network is regarded as a challenge by researchers. This issue represents serious problems for many protocols particularly those that depend on sending messages as…

Genetic Programming (GP) is a computationally intensive technique which also has a high degree of natural parallelism. Parallel computing architectures have become commonplace especially with regards Graphics Processing Units (GPU). Hence,…

Parallel algorithms on CPU and GPU are implemented for the Unified Gas-Kinetic Scheme and their performances are investigated and compared by a two dimensional channel flow case. The parallel CPU algorithm has a one dimensional block…

Nowadays, the bulk of Internet traffic uses TCP protocol for reliable transmission. But the standard TCP's performance is very poor in High Speed Networks (HSN) and hence the core gigabytes links are usually underutilization. This problem…

Modern database clusters entail two levels of networks: connecting CPUs and NUMA regions inside a single server in the small and multiple servers in the large. The huge performance gap between these two types of networks used to slow down…

As the data size in Machine Learning fields grows exponentially, it is inevitable to accelerate the computation by utilizing the ever-growing large number of available cores provided by high-performance computing hardware. However, existing…

The Grid Datafarm architecture is designed for global petascale data-intensive computing. It provides a global parallel filesystem with online petascale storage, scalable I/O bandwidth, and scalable parallel processing, and it can exploit…

Dynamic Parallelism (DP) is a runtime feature of the GPU programming model that allows GPU threads to execute additional GPU kernels, recursively. Apart from making the programming of parallel hierarchical patterns easier, DP can also…

For ultra-reliable, low-latency communications (URLLC) applications such as mission-critical industrial control and extended reality (XR), it is important to ensure the communication quality of individual packets. Prior studies have…