CAJ | 학술논문

提出一种应用于网络处理器的Hash算法,通过建立新型查找表的结构和构造两级Hash函数,能够有效地解决Hash冲突的问题。描述Hash表的软件建立流程和硬件查找过程,在Hash查找的基础上,给出硬件表项的学习过程和老化方法,简化表项的更新操作。针对不同的应用,建立不同类型的Hash表,合理地利用内外部存储资源,兼顾了存储资源和处理速度的平衡。实验结果表明,该算法对各种查找表中不同的表项数目和关键词长度均具有较好的兼容性,成功查找的平均长度为2,减少了存储器的访存次数,其单个微引擎的查找速度高达25 Mb/s,能够满足网络处理器接口处理带宽20 Gb/s的要求。
제출일충응용우망락처리기적Hash산법,통과건립신형사조표적결구화구조량급Hash함수,능구유효지해결Hash충돌적문제。묘술Hash표적연건건립류정화경건사조과정,재Hash사조적기출상,급출경건표항적학습과정화노화방법,간화표항적경신조작。침대불동적응용,건립불동류형적Hash표,합리지이용내외부존저자원,겸고료존저자원화처리속도적평형。실험결과표명,해산법대각충사조표중불동적표항수목화관건사장도균구유교호적겸용성,성공사조적평균장도위2,감소료존저기적방존차수,기단개미인경적사조속도고체25 Mb/s,능구만족망락처리기접구처리대관20 Gb/s적요구。
A novel Hash algorithm is proposed in this paper for network processor application. It resolves Hash collision problem by constructing new look up table and new two-level Hash function. The software processing and hardware lookup flow of Hash table are descripted, and the learning process and ageing machine for entry of table are designed for simplifying the entry updating operation. For different engineering applications,the algorithm sets up different Hash table, which makes the efficience of memory utilization improved and the tradeoff between memory and processing speed optimized. Simulation results show the algorithm works well despite of the number of table entry and the size of keyword. The average length of look up’ s success is 2 and the memory access times is reduced dramaticlly. The look up speed of micro-engine is improved to 25 Mb/s,satisfing the requinrement of 20 Gb/s bandwidth performance of network processor.