ScholarLens Back to search
2015•Unpublished venueRequires access

LW-RDMA

Bongjae Kim, Hyedong Jung

Open publisher page 0 citations

Abstract

Due to the development of low-latency networks such as InfiniBand and Myrinet, cluster-based computing systems are now commonly used when implementing computing-intensive applications. InfiniBand is the de-facto standard for high-performance cluster-based computing systems. InfiniBand supports RDMA (Remote Direct Memory Access) features and provides a native API for these features. The use of the native-API-like VPI Verbs API is not easy owing to its complexity. However, customizing and optimizing can be done to improve the development efficiency. In this paper, we design and implement a lightweight RDMA API to improve the usability and applicability of InfiniBand-based clusters. The proposed method is termed LW-RDMA. The LW-RDMA API is lightweight and simple to use by those without expertise in the area of InfiniBand architecture and/or related details. In the experimental results, the implemented LW-RDMA showed better data transfer efficiency performance in some cases when compared to other RDMA techniques.

About this research paper

What this paper is about

Due to the development of low-latency networks such as InfiniBand and Myrinet, cluster-based computing systems are now commonly used when implementing computing-intensive applications. InfiniBand is the de-facto standard for high-performance cluster-based computing systems. InfiniBand supports RDMA (Remote Direct Memory Access) features and provides a native API for these features. The use of the native-API-like VPI Verbs API is not easy owing to its complexity. However, customizing and optimizing can be done to improve the development efficiency. In this paper, we design and implement a lightweight RDMA API to improve the usability and applicability of InfiniBand-based clusters. The proposed method is termed LW-RDMA. The LW-RDMA API is lightweight and simple to use by those without expertise in the area of InfiniBand architecture and/or related details. In the experimental results, the implemented LW-RDMA showed better data transfer efficiency performance in some cases when compared to other RDMA techniques.

Why it matters

A significance statement is not available in the OpenAlex record.

Key contribution

A contribution statement is not available in the OpenAlex record.

Method / approach

Method details are not available in the OpenAlex metadata.

Main findings

Findings are not separately available in the OpenAlex metadata.

Limitations

Limitations are not available in the OpenAlex metadata.

Applications

Application details are not available in the OpenAlex metadata.

Available abstract

Due to the development of low-latency networks such as InfiniBand and Myrinet, cluster-based computing systems are now commonly used when implementing computing-intensive applications. InfiniBand is the de-facto standard for high-performance cluster-based computing systems. InfiniBand supports RDMA (Remote Direct Memory Access) features and provides a native API for these features. The use of the native-API-like VPI Verbs API is not easy owing to its complexity. However, customizing and optimizing can be done to improve the development efficiency. In this paper, we design and implement a lightweight RDMA API to improve the usability and applicability of InfiniBand-based clusters. The proposed method is termed LW-RDMA. The LW-RDMA API is lightweight and simple to use by those without expertise in the area of InfiniBand architecture and/or related details. In the experimental results, the implemented LW-RDMA showed better data transfer efficiency performance in some cases when compared to other RDMA techniques.

Key concepts: InfiniBand, Remote direct memory access, Myrinet, Computer science, Operating system, Low latency (capital markets), Message passing, Parallel computing

Related papers

Related paper 1Related paper 2
Back to paper searchBrowse research topicsOriginal source
LW-RDMA — Research Paper | ScholarLens