2011•Murdoch Research Repository (Murdoch University)Requires access

Design and analysis of using a programmable network interface for high speed networks beyond the 10 Gbps

Mohamed Elbeshti, Mike Dixon, Terry W. Koziniec

Open publisher page 0 citations

Abstract

The networks speed has been advancing rapidly in providing higher transmission rate 10 Gbps and over. These improvements based on the demand of enhancing the network services, improving their bandwidth and integrating advanced technology. As the speed of networks exceeds 10 Gbps, the design and implementation of high-performance Network Interfaces (NI) for the current and the Next Generation Network (NGN) server applications that employ TCP/IP and UDP/IP as the communication protocol of choice is becoming very challenging. Using the General Purpose Processor (GPP) as a main core processor in the NI to offload the TCP/IP or UDP/IP functions, can deliver some important features to NI such as scalability and short developing time. However, it is not clear that using GPP can support the new speed line over the 10 Gbps. Furthermore, it is necessary to find out the clock rate Hz limit of these GPP in supporting the processing of NIs. In this research, we have proposed a new programmable Ethernet NI (ENI) model design to support the high speed transmission. This model supports the Large Segment Offload (LSO) for sending side and a novel algorithm for receiving side called Receiving Side Amalgamating Algorithm (RSAA). As a result, a 240 MHz RISC core can be used in this Ethernet NI card for a wide range of the transmission line speeds up to 100 Gbps when a jumbo packet assigned as a default size for a network.

About this research paper

What this paper is about

The networks speed has been advancing rapidly in providing higher transmission rate 10 Gbps and over. These improvements based on the demand of enhancing the network services, improving their bandwidth and integrating advanced technology. As the speed of networks exceeds 10 Gbps, the design and implementation of high-performance Network Interfaces (NI) for the current and the Next Generation Network (NGN) server applications that employ TCP/IP and UDP/IP as the communication protocol of choice is becoming very challenging. Using the General Purpose Processor (GPP) as a main core processor in the NI to offload the TCP/IP or UDP/IP functions, can deliver some important features to NI such as scalability and short developing time. However, it is not clear that using GPP can support the new speed line over the 10 Gbps. Furthermore, it is necessary to find out the clock rate Hz limit of these GPP in supporting the processing of NIs. In this research, we have proposed a new programmable Ethernet NI (ENI) model design to support the high speed transmission. This model supports the Large Segment Offload (LSO) for sending side and a novel algorithm for receiving side called Receiving Side Amalgamating Algorithm (RSAA). As a result, a 240 MHz RISC core can be used in this Ethernet NI card for a wide range of the transmission line speeds up to 100 Gbps when a jumbo packet assigned as a default size for a network.

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

The networks speed has been advancing rapidly in providing higher transmission rate 10 Gbps and over. These improvements based on the demand of enhancing the network services, improving their bandwidth and integrating advanced technology. As the speed of networks exceeds 10 Gbps, the design and implementation of high-performance Network Interfaces (NI) for the current and the Next Generation Network (NGN) server applications that employ TCP/IP and UDP/IP as the communication protocol of choice is becoming very challenging. Using the General Purpose Processor (GPP) as a main core processor in the NI to offload the TCP/IP or UDP/IP functions, can deliver some important features to NI such as scalability and short developing time. However, it is not clear that using GPP can support the new speed line over the 10 Gbps. Furthermore, it is necessary to find out the clock rate Hz limit of these GPP in supporting the processing of NIs. In this research, we have proposed a new programmable Ethernet NI (ENI) model design to support the high speed transmission. This model supports the Large Segment Offload (LSO) for sending side and a novel algorithm for receiving side called Receiving Side Amalgamating Algorithm (RSAA). As a result, a 240 MHz RISC core can be used in this Ethernet NI card for a wide range of the transmission line speeds up to 100 Gbps when a jumbo packet assigned as a default size for a network.

Key concepts: Computer science, Network interface controller, Computer network, Ethernet, Network interface, Network processor, Scalability, Network packet

Related papers

Back to paper searchBrowse research topicsOriginal source
Design and analysis of using a programmable network interface for high speed networks beyond the 10 Gbps — Research Paper | ScholarLens