本文選題：并行協(xié)議棧 + 復(fù)制技術(shù)��； 參考：《西安電子科技大學(xué)》2014年碩士論文

【摘要】：隨著網(wǎng)絡(luò)通信技術(shù)的快速發(fā)展，網(wǎng)絡(luò)帶寬和流量不斷增大，，傳統(tǒng)單核框架下的網(wǎng)絡(luò)協(xié)議棧的處理能力很容易成為系統(tǒng)性能的瓶頸。多核處理器框架以及并行計算等技術(shù)可以有效解決單個核處理器能力不夠的問題，但是傳統(tǒng)的網(wǎng)絡(luò)協(xié)議棧不適應(yīng)多核框架。數(shù)據(jù)包級別的并行協(xié)議棧是一種多核框架下的協(xié)議棧，具有并行粒度細(xì)，可提高單個連接性能等優(yōu)點，但存在多進(jìn)程競爭協(xié)議狀態(tài)的問題。 本文針對數(shù)據(jù)包級別并行協(xié)議棧中多線程競爭協(xié)議狀態(tài)的問題，在深入研究復(fù)制技術(shù)的基礎(chǔ)上，給出了一種基于復(fù)制技術(shù)的協(xié)議棧并行化方案。在方案設(shè)計中，運用復(fù)制技術(shù)，使得協(xié)議狀態(tài)被復(fù)制到各個線程中，避免了數(shù)據(jù)競爭；采用了基于原子操作的無鎖算法改進(jìn)了消息隊列操作，提高了多個線程訪問消息隊列的性能；且使用了零拷貝技術(shù)打通了網(wǎng)卡和用戶層之間的通道，減少了中斷次數(shù)以及數(shù)據(jù)包復(fù)制次數(shù)，從而提高了底層數(shù)據(jù)包傳輸模塊的性能�；谝陨详P(guān)鍵技術(shù)，在用戶態(tài)實現(xiàn)了基于復(fù)制技術(shù)的并行協(xié)議棧。實驗結(jié)果表明，該協(xié)議棧性能優(yōu)于數(shù)據(jù)包級別并行協(xié)議棧，有效地解決了鎖競爭問題。
[Abstract]:With the rapid development of network communication technology, the network bandwidth and traffic are increasing, and the processing ability of the network protocol stack under the traditional single-core framework is easy to become the bottleneck of the system performance. Multi-core processor framework and parallel computing technology can effectively solve the problem of single core processor capability, but the traditional network protocol stack is not suitable for multi-core framework. Packet level parallel protocol stack is a multi-core protocol stack with fine parallel granularity and can improve the performance of a single connection. However there is a problem of multi-process contention protocol state. In this paper, aiming at the problem of multithread competing protocol status in packet level parallel protocol stack, a protocol stack parallelization scheme based on replication technology is proposed on the basis of in-depth study of replication technology. In the scheme design, the protocol state is copied to every thread by using replication technology, and the lock free algorithm based on atomic operation is used to improve the message queue operation. The performance of multiple threads accessing message queue is improved, and the channel between network card and user layer is opened by using zero-copy technology, which reduces the number of interrupts and the number of packet replication, thus improving the performance of the underlying data packet transmission module. Based on the above key technologies, the parallel protocol stack based on replication technology is implemented in user state. The experimental results show that the performance of the protocol stack is superior to that of the packet level parallel protocol stack, and the lock competition problem is effectively solved.
【學(xué)位授予單位】：西安電子科技大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2014
【分類號】：TP393.04

【參考文獻(xiàn)】

相關(guān)期刊論文 前5條

1 可向民,龔正虎,夏建東;零拷貝技術(shù)及其實現(xiàn)的研究[J];計算機工程與科學(xué);2000年05期

2 杜凱;繆嘉嘉;楊樹強;王懷民;;數(shù)據(jù)庫復(fù)制技術(shù)研究進(jìn)展[J];計算機工程與科學(xué);2008年07期

3 李晉,葛敬國;Linux下互斥機制及其分析[J];計算機應(yīng)用研究;2005年08期

4 張宇雷;黃皓;;基于網(wǎng)絡(luò)處理器的零拷貝技術(shù)[J];計算機應(yīng)用研究;2007年01期

5 夏高;劉斌;;用于高速網(wǎng)絡(luò)入侵檢測系統(tǒng)的并行TCP/IP協(xié)議棧[J];清華大學(xué)學(xué)報(自然科學(xué)版);2011年07期

相關(guān)博士學(xué)位論文 前1條

1 李波;基于異構(gòu)多核平臺的優(yōu)化編程研究[D];華中科技大學(xué);2011年

本文編號：2048520