| dc.description.abstract | number of hardware appliances. Furthermore, deploying a new Network Function(NF) often requires installing yet another variety of proprietary hardware. This iscompounded by the increasing costs of energy and capital investment. The conceptof Network Function Virtualization (NFV) originated from the requirement of theservice providers to enhance performance, reduce deployment and operating cost,foster competition, and improve scalability and manageability. NFV takes advantageof cloud computing and virtualization technology to accelerate the deployment ofVirtualized Network Functions (VNFs). NFV promises to bring a multitude ofbenefits to the network field as it decouples software from hardware which meansthat the implementation of both hardware and software is no longer dependent oneach other. In other words, NFV promotes the implementation of network functionsin software instead of installing new dedicated hardware every time a new networkfunction is added. The service provider can simply launch a new VNF which canrun on a commodity hardware and can be managed without the need of modifyingthe physical infrastructure. Since many VNFs prototype are available to be studiedand analyzed, it is equally important to identify their performance, and ability tomeet the user expectations.
In this project, we study the performance of VNF in the context of a virtualizedIP Multimedia Subsystem (IMS), called ClearWater, which is an open source NFVbased IMS. Clearwater follows IMS architectural principles and supports all of theservices, and standardized interfaces expected of an IMS. ClearWater is available to bedeployed in two widely used virtualization technologies: an open source Linux-basedcontainer (Docker), and Virtual Machine (VM). Similar to typical IMS, ClearWaterutilizes Session Initiation Protocol (SIP) for call session establishment and control.SIP is a signalling protocol used for managing multimedia sessions over an IP datanetwork, and negotiating the parameters of the created session.
To this end, we consider a common deployment where all the VNF services arerunning on the same physical machine. The main goal is to study and analyze theinitial user registration process of ClearWater for both deployment options, and toderive an analytical model based on queuing networks. The experimental results willform the basis to eventually refine and validate the analytic model. The project aimsto discover the influence of different system features like resource allocation, servicerequest arrival rates, number of processing instances on the system performance interms of response time, successful and failure rate, and resource utilization. | |
