Akademik Veri Yönetim Sistemi
Tezin Türü: Yüksek Lisans
Tezin Yürütüldüğü Kurum: Middlesex University, School of Engineering and Information Sciences, Computer Communications , İngiltere
Tezin Onay Tarihi: 2011
Tezin Dili: Türkçe
Öğrenci: Ravi Luchmun
Danışman: Enver Ever
Özet:
This thesis presents simulation studies together with
numerical results for the performance evaluation of systems such as single
server with virtualisation, single server with virtualisation including
priority queuing and finite queuing capacity for two queues, as well as
performability of multiple homogenous servers with failures. Simulation results obtained have been
compared to published results of analytical studies in the literature and the
accuracy of the simulation has been analysed in turn. An application of the
virtualised single server with priority and finite queuing is that of a traffic
centre or router which has to deal with both voice and data traffic. Since voice traffic is delay sensitive in
nature, it has been modelled using the priority and finite queue.
The methodology used in this thesis has been of a building
block approach. First an M/M/1/∞/∞/FCFS
model has been implemented and performance analysis results obtained and
validated. Once the single server simulation is validated, M/M/c/∞/∞/FCFS model
is considered and validated as well.
From there, an M/M/1 model with two queues, one for voice and another
for data, representing virtualised servers has been developed. It is worth noting that components of a
system do fail at some point during their lifetime and hence pure performance
analysis is not representative of the real world. Therefore, performability simulations for
M/M/c systems have also been implemented, validated and presented.
In this work, server virtualisation with priority queuing
and finite queuing capacity simulations has also been implemented. It is always a zero-sum game, for any system
offering priority or quality of service, where one suffers performance issues
at the expense of another. From
simulation results obtained, we show that if voice traffic is prioritised over
data traffic, the latter experiences longer waiting times.