公开(公告)号：US10084661B2

公开(公告)日：2018-09-25

申请号：US15416035

申请日：2017-01-26

Applicant: Macau University of Science and Technology

Inventor： Qinglin Zhao ,  Rongchang Duan ,  Hanwen Zhang ,  Yujun Zhang ,  Zhongcheng Li

IPC: H04W24/00 ,  H04W84/18 ,  H04L12/24 ,  H04W4/00 ,  H04L12/26 ,  H04W4/70

CPC classification number: H04L41/142 ,  H04L43/08 ,  H04L43/0888 ,  H04L67/12 ,  H04W4/70

Abstract: A method for evaluating performance of a data communication network with a receiver node arranged to initiate data transmission thereto from a plurality of transmitter nodes directly connected therewith includes applying to the data communication network a stochastic operation model arranged to model operation characteristics of the receiver node, and determining throughput or power consumption of the network based on the stochastic operation model. The operation characteristics include: a duration of a work cycle of the receiver node (Tcycle), the work cycle including an active data communication period (Thold) and an inactive period (Tdwell); a duration of the active data communication period (Thold), the active data communication period including a number of data communication events; the number of data communication events in the work cycle n; a duration of the respective data communication events (Ttx); and amount of data successfully received at the receiver node in the work cycle (Bi).

公开(公告)号：US20180212836A1

公开(公告)日：2018-07-26

申请号：US15416035

申请日：2017-01-26

Applicant: Macau University of Science and Technology

Inventor： Qinglin Zhao ,  Rongchang Duan ,  Hanwen Zhang ,  Yujun Zhang ,  Zhongcheng Li

IPC: H04L12/24 ,  H04W4/00 ,  H04L12/26

CPC classification number: H04L41/142 ,  H04L43/08 ,  H04L43/0888 ,  H04W4/70

Abstract: A method for evaluating performance of a data communication network with a receiver node arranged to initiate data transmission thereto from a plurality of transmitter nodes directly connected therewith includes applying to the data communication network a stochastic operation model arranged to model operation characteristics of the receiver node, and determining throughput or power consumption of the network based on the stochastic operation model. The operation characteristics include: a duration of a work cycle of the receiver node (Tcycle), the work cycle including an active data communication period (Thold) and an inactive period (Tdwell); a duration of the active data communication period (Thold), the active data communication period including a number of data communication events; the number of data communication events in the work cycle n; a duration of the respective data communication events (Ttx); and amount of data successfully received at the receiver node in the work cycle (Bi).

公开(公告)号：US10001930B2

公开(公告)日：2018-06-19

申请号：US14972047

申请日：2015-12-16

Applicant: Macau University of Science and Technology

Inventor： Li Feng ,  Liang Zhou ,  Zhijun Xu ,  Yujun Zhang

IPC: G06F3/06

CPC classification number: G06F3/0613 ,  G06F3/0659 ,  G06F3/068 ,  G06F13/00 ,  H04L47/50 ,  H04L69/12 ,  H04L69/321

Abstract: A method for implementing a packet I/O engine on a programmable computing platform is provided, where the engine performs I/O functions for plural threads generated by a plurality of user applications. In the method, the platform is configured such that only one thread is permitted to initialize and configure the resources. Furthermore, I/O-device queues each for buffering packets either transmitted to or received from an individual external I/O device are set up. For a plurality of unsafe I/O-device queues determined, among the I/O-device queues, to be multi-thread unsafe, a plurality of multi-producer, multi-consumer software queues for buffering packets delivered between the plurality of the unsafe I/O-device queues and the plurality of user applications is set up. In particular, the plurality of multi-producer, multi-consumer software queues is configured such that the unsafe I/O-device queues are collectively synchronized to maintain data consistency in packet delivery in the presence of multiple threads.

公开(公告)号：US09961002B2

公开(公告)日：2018-05-01

申请号：US14972062

申请日：2015-12-16

Applicant: Macau University of Science and Technology

Inventor： Li Feng ,  Liang Zhou ,  Zhijun Xu ,  Yujun Zhang

IPC: H04L12/803 ,  H04L12/707 ,  H04L12/851 ,  H04L29/06

CPC classification number: H04L47/125 ,  H04L45/24 ,  H04L47/2483 ,  H04L69/22

Abstract: Compute-intensive packet processing (CIPP) in a computer system comprising a programmable computing platform is accelerated by using a packet I/O engine, implemented on the platform, to perform packet I/O functions, where the packet I/O engine is configured to achieve direct access to a network interface card (NIC) from a user application. For a Linux-based computer system, standard I/O mechanisms of Linux are bypassed and only the packet I/O engine is used in performing the I/O functions. Furthermore, the computer system is configured to: process a batch of packets, instead of packet by packet, in every function call; and when moving a packet between a buffer of an individual user application and a queue of the packet I/O engine, copy a packet descriptor of the packet instead the entire packet. In addition, workflows across different working threads are balanced and parallelism is exploited to fully utilize resources of the platform.

公开(公告)号：US20170163538A1

公开(公告)日：2017-06-08

申请号：US14972062

申请日：2015-12-16

Applicant: Macau University of Science and Technology

Inventor： Li Feng ,  Liang Zhou ,  Zhijun Xu ,  Yujun Zhang

IPC: H04L12/803 ,  H04L12/851 ,  H04L12/707 ,  H04L29/06

CPC classification number: H04L47/125 ,  H04L45/24 ,  H04L47/2483 ,  H04L69/22

Abstract: Compute-intensive packet processing (CIPP) in a computer system comprising a programmable computing platform is accelerated by using a packet I/O engine, implemented on the platform, to perform packet I/O functions, where the packet I/O engine is configured to achieve direct access to a network interface card (NIC) from a user application. For a Linux-based computer system, standard I/O mechanisms of Linux are bypassed and only the packet I/O engine is used in performing the I/O functions. Furthermore, the computer system is configured to: process a batch of packets, instead of packet by packet, in every function call; and when moving a packet between a buffer of an individual user application and a queue of the packet I/O engine, copy a packet descriptor of the packet instead the entire packet. In addition, workflows across different working threads are balanced and parallelism is exploited to fully utilize resources of the platform.

公开(公告)号：US20170160954A1

公开(公告)日：2017-06-08

申请号：US14972047

申请日：2015-12-16

Applicant: Macau University of Science and Technology

Inventor： Li Feng ,  Liang Zhou ,  Zhijun Xu ,  Yujun Zhang

IPC: G06F3/06

CPC classification number: G06F3/0613 ,  G06F3/0659 ,  G06F3/068 ,  G06F13/00 ,  H04L47/50 ,  H04L69/12 ,  H04L69/321

Abstract: A method for implementing a packet I/O engine on a programmable computing platform is provided, where the engine performs I/O functions for plural threads generated by a plurality of user applications. In the method, the platform is configured such that only one thread is permitted to initialize and configure the resources. Furthermore, I/O-device queues each for buffering packets either transmitted to or received from an individual external I/O device are set up. For a plurality of unsafe I/O-device queues determined, among the I/O-device queues, to be multi-thread unsafe, a plurality of multi-producer, multi-consumer software queues for buffering packets delivered between the plurality of the unsafe I/O-device queues and the plurality of user applications is set up. In particular, the plurality of multi-producer, multi-consumer software queues is configured such that the unsafe I/O-device queues are collectively synchronized to maintain data consistency in packet delivery in the presence of multiple threads.