Date of Award
Dissertation (Access Restricted)
Doctor of Philosophy
Sajjan G Shiva
Software-Defined Networks (SDNs) have become an important infrastructure inmany network-intensive applications that require high bandwidths for big data transferover long distances. They enable the network control plane to be decoupled from thedata plane and assign the control to a programmable software unit, i.e. controller. Withsuch a logically centralized control, SDNs provide the capability of both advance andimmediate bandwidth reservations, the former reserving bandwidth ahead of time in afuture time slot for a known duration, while the latter allocating bandwidth uponavailability in the next immediate time slot. An ongoing data transfer task based on animmediate reservation (IR) may be preempted by the activation of an advancereservation (AR) due to the lack of bandwidth. The preemption of IRs, also termed asIR-preemption problem, is largely due to the neglect of the presence of IRs at the timeof scheduling an incoming AR. In SDNs with multiple logically centralized controllers,it is challenging to maintain accurate link-state information, perceived as a GlobalNetwork View (GNV), at every controller in a consistent manner. Consequently,bandwidth scheduling in such environments may lead to blocking or rejection ofreservation requests, also termed as AR-blocking problem, which deteriorates as thelevel of inaccuracy/inconsistency increases. In order to minimize the service disruptioncaused by IR-preemption and AR-blocking problems, and to ensure optimal networkutilization, we propose a class of bandwidth reservation solutions: (i) We design ascheduling algorithm to route incoming ARs with QoS constraints, such as bandwidth,delay, deadline, etc., with minimal impact on ongoing IRs. (ii) We design a run-timepreemption scheme to minimize the actual number of IRs that must be preempted atthe activation of an AR. (iii) We formulate advance bandwidth scheduling in SDNs asan optimization problem and propose a randomization-based routing scheme toschedule AR requests such that the total number of blocked reservations due to theinaccurate/inconsistent GNV is minimized. The performance superiority of theproposed bandwidth reservation solutions is illustrated by extensive simulations incomparison with existing methods.
Dissertation or thesis originally submitted to the local University of Memphis Electronic Theses & dissertation (ETD) Repository.
Dharam, Poonam, "QoS Routing for Big Data Transfer in Software-Defined Networks" (2015). Electronic Theses and Dissertations. 2274.