存储区域网中光纤通道和iSCSI协议的性能分析

在存储环境中,SCSI协议是被广泛采用的块存储管理协议。不同的传输机制,如,并行SCSI、光纤通道和iSCSI(internetSCSI)都可以被用于传输SCSI协议数据。对于并行SCSI来说,传输距离短、连接的设备数目少是它的主要缺点。FC和iSCSI是下一代SCSI传输机制的主要竞争对手。通过对FC和iSCSI两个协议之间不同方面的分析,评价FC和iSCSI是否满足吉比特速率、可扩展互连设备和满足存储环境和存储协议的需要。     

Throughput-delay performance of packet-switching multiple-access channel with power capture

Power capture is the ability of a receiver to receive correctly the strongest of several messages that arrive during overlapping intervals. The presence of power capture in multiple-access packet switching channels introduces inequality in the access conditions to the channel when packets arrive at the receiver at different power levels. Two models for such channels under slotted ALOHA protocol with power capture are considered. In the first, a packet is transmitted at one of K possible power levels, and it is received correctly if no other packet is transmitted at the same or higher level. The second model is of a ground radio network where a received packet's power is inversely proportional to the distance it propagates. In this model, a packet transmitted from a distance r is received correctly if no other packet is transmitted from a distance smaller than a · r (a ? 1). A technique developed by Lam (1974) is generalized to obtain the throughput-delay characteristics of the channel for both models. The effect of high power traffic on the lower power traffic is discussed, and it is shown that the ‘Sisyphus distance’ phenomenon predicted by Abramson (1977) for the case a = 1 in the second model does not exist in the more realistic case of a > 1.     

On the analysis of a symmetric polling system with single-message buffers

A performance analysis is provided for a polling system consisting of statistically identical stations with single-message buffers and Poisson arrival streams. Switchover and message service times are assumed to be generally distributed. Some errors in the past analysis are pointed out. We express such performance measures as the mean polling cycle time, the mean message response time, and the mean number of messages at an arbitrary time in terms of the mean number of massages served in a polling cycle. Our mean message response time reduces to that for an FCFS M/G/1//N queue (machine interference model) in the limit of zero switchover time.