首先讨论和总结了电子邮件系统的基本理论、工作原理和相关协议。通过对现有的服务器内部架构的分析和比较,最后提出了基于进程池的SMTP服务器、POP3服务器、WMG服务器等模块设计。通过对系统的功能和性能测试,表明该服务器实现了设计目标并具有较高的性能。

缓存失效报告法是在移动计算环境中普遍采用的缓存方案，即服务器定期广播缓存失效报告．该报告中包含了最近被更新的对象-根据报告内容，客户可使缓存中被更新对象失效．以保证缓存的有效性。缓存失效报告技术可采用多种策略，本文在研究多种缓存策略的基础上，提出了一种针对缓存失效方案的归类研究法，可对具体策略加以分析和评估．在此基础上对其性能进行改良，以形成更适应特定移动计算环境要求的缓存方案。

To solve the problem that most of existing layered multicast protocols cannot adapt to dynamic network conditions because their layers are coarsely granulated and static, a new congestion control mechanism for dynamic adaptive layered multicast(DALM) is presented. In this mechanism, a novel feedback aggregating algorithm is put forward, which can dynamically determine the number of layers and the rate of each layer, and can efficiently improve network bandwidth utilization ratio. Additionally, because all layers is transmitted in only one group, the intricate and time-consuming internet group management protocol(IGMP) operations, caused by receiver joining a new layer or leaving the topmost subscribed layer, are thoroughly eliminated. And this mechanism also avoids other problems resulted from mtdtiple groups. Simulation results show that DALM is adaptive and TCP friendly.

Congestion control mechanism for multicast is one of the key problems of designing multicast protocol. In this paper, a congestion control mechanism for multicast is presented: RBMCC, which based rate and initiated by both sender and receiver. RBMCC uses three factors (loss rate of receivers, aggregated ACK of intermediate nodes and TCP-friendly rate computed by sender) to provide timely and accurate response on behalf of congestion. Experiment results show that RBMCC is scaleable and TCP-friendly.

分析了传统磁盘队列的存储管理开销和读写性能，针对磁盘队列I/O已成为影响消息服务器性能的首要瓶颈，提出了一种高效磁盘队列I/O机制-FlashQ。FlashQ采用物理上连续的磁盘块作为磁盘队列，以紧凑方式组织数据，采用延迟写策略和预先读策略提高读写性能。实验表明，FlashQ的性能比传统磁盘队列好。