The Performance Analysis and Improvement of E-mail Servers

Y.-W. Bai and C.-J. Cheng (Taiwan)


E-mail Server, System Response Time, Queueing Model, Buffer Full Probability.


Due to the increasing size of files attached to e-mails, the working load of e-mail servers can be very heavy. Thus, in this paper, we propose a queueing model both to analyze the performance of e-mail servers and to use the queueing network simulation tool (QNAT) to simulate the operational characteristics of e-mail servers. Both methods and our measurements show that when the average data arrival rate is increased, the system response time and the probability of an e-mail buffer full will increase. Therefore, we also propose a couple of strategies to reduce the system response time and the buffer full probability. These strategies include (1) increasing CPU speed, (2) using multiple CPUs, and (3) increasing buffer size. In addition, to reduce the buffer full probability, we increase the buffer size. Overall, from analyses and simulation results of the proposed strategies, we learn an e-mail server that uses multiple CPUs or that uses faster CPUs will have an improvement in the system response time and the buffer full probability, due to the increase of the service speed.

Important Links:

Go Back