The conventional workstation may not be sufficient enough for the bandwidth provided by the high-speed networks due to the performance bottleneck in the I/O subsystem. Specifically, the “delayed acknowledgment” and “windowing out” problems may severely damage system performance. In this work, we propose a novel I/O subsystem architectural design with the hierarchical I/O memories (HIOM) to address these problems and to support the highly pipelined data transmissions in high-speed network environment. The HIOM architecture effectively reduces the data blocking by moving data between the different levels of the I/O buffers. It also provides large receiving I/O buffer space to offset the effect of network propagation delay during the long-dist...
This book provides an overview of recent advances in memory interface design at both the architectur...
Achieving high-speed network I/O on distributedmemory systems is a hard problem because their archit...
The state-of-the-art design methodology for high-speed I/O links is to specify component-level desig...
The conventional workstation may not be sufficient enough for the bandwidth provided by the high-spe...
This paper looks at the I/O bottleneck in operating systems, with particular focus on high-speed net...
This paper describes a new host interface architecture for high-speed networks operating at 800 of M...
Fast non-volatile memories are exposing inefficiencies in traditional I/O stacks. Though there have ...
The advent of high-speed networks may soon increase the network bandwidth available to workstation c...
Abstract. The technology advances made in supercomputers and high performance computing clusters ove...
There are two complementary trends in the computer and communications fields. Increasing processor p...
Abstract—Current leadership-class machines suffer from a significant imbalance between their computa...
The increasing number of cores per node has propelled the performance of leadershipscale systems fro...
Though input/output (I/O) from mass storage continues to be a bottleneck in current generation distr...
Introduction Applications are an important driving force behind the emergence of new machine archit...
High performance computing (HPC) has crossed the Petaflop mark and is reaching the Exaflop range qui...
This book provides an overview of recent advances in memory interface design at both the architectur...
Achieving high-speed network I/O on distributedmemory systems is a hard problem because their archit...
The state-of-the-art design methodology for high-speed I/O links is to specify component-level desig...
The conventional workstation may not be sufficient enough for the bandwidth provided by the high-spe...
This paper looks at the I/O bottleneck in operating systems, with particular focus on high-speed net...
This paper describes a new host interface architecture for high-speed networks operating at 800 of M...
Fast non-volatile memories are exposing inefficiencies in traditional I/O stacks. Though there have ...
The advent of high-speed networks may soon increase the network bandwidth available to workstation c...
Abstract. The technology advances made in supercomputers and high performance computing clusters ove...
There are two complementary trends in the computer and communications fields. Increasing processor p...
Abstract—Current leadership-class machines suffer from a significant imbalance between their computa...
The increasing number of cores per node has propelled the performance of leadershipscale systems fro...
Though input/output (I/O) from mass storage continues to be a bottleneck in current generation distr...
Introduction Applications are an important driving force behind the emergence of new machine archit...
High performance computing (HPC) has crossed the Petaflop mark and is reaching the Exaflop range qui...
This book provides an overview of recent advances in memory interface design at both the architectur...
Achieving high-speed network I/O on distributedmemory systems is a hard problem because their archit...
The state-of-the-art design methodology for high-speed I/O links is to specify component-level desig...