Posters on Theme 5

THEME 5 APPLICATION PLATFORM TESTBED

Posters
“Virtualized Reconfigurable Hardware in the SAVI Testbed” Stuart Byma, Hadi Bannazadeh, J. Gregory Steffan, Alberto Leon-Garcia, Paul Chow University of Toronto
“Real-time Enhancement of IMS Quality of Service using SAVI SDI” Lilin Zhang, Jieyu Lin, Hadi Bannazadeh, Alberto Leon-Garcia, University of Toronto
“An Orchestration Service of SAVI Testbed; A Heat Approach” Honbin Lu, Hadi Bannazadeh, Alberto Leon-Garcia, University of Toronto
“Big Data as a Service for SAVI Testbed” Honbin Lu, Hadi Bannazadeh, Alberto Leon-Garcia, University of Toronto
“Cloud-RAN on SAVI; a GSM approach” Mohammad-Sina Tavoosi-Monfared, Hadi Bannazadeh, Alberto Leon-Garcia, University of Toronto
“Monitoring and Measurement as a Service in SDI deployed on SAVI testbed” Jie Yu (Eric) Lin, Rajsimman Ravichandiran, Hadi Bannazadeh, Alberto Leon-Garcia, University of Toronto
“Efficient Multicast algorithm on SAVI network” Sai Qian Zhang, Hadi Bannazadeh, Alberto Leon-Garcia, University of Toronto
“End-to-End Traffic Control in the SAVI Testbed” Thomas Lin, Hadi Bannazadeh, Alberto Leon-Garcia, University of Toronto
“L2/L3 Overlay Software Defined Network on SAVI Testbed” Khashayar Hoseinzadeh, Hadi Bannazadeh, Alberto Leon-Garcia, University of Toronto

POSTER & DEMONSTRATION:

Virtualized Reconfigurable Hardware in the SAVI Testbed

Stuart Byma, Hadi Bannazadeh, J. Gregory Steffan, Alberto Leon-Garcia, Paul Chow University of Toronto

This poster presents a novel approach for integrating virtualized FPGA-based hardware accelerators into the SAVI testbed, with minimal virtualization overhead. Partially reconfigurable regions inside FPGAs are offered as generic cloud resources through the testbed OpenStack management system, thereby allowing users to “boot” custom designed or predefined network-connected hardware accelerators with the same commands they would use to boot a regular Virtual Machine. The poster details the hardware and software framework to enable this virtualization, and also shows two example use cases — a custom UDP load balancer and an application used to give OpenFlow the ability to execute matching and actions on VXLAN tunnelled traffic. The latter application will be showcased in a demonstration.

Real-time Enhancement of IMS Quality of Service using SAVI SDI

Lilin Zhang, Jieyu Lin, Hadi Bannazadeh, Alberto Leon-Garcia, University of Toronto

Traditional IMS/VoIP services lack real-time bandwidth provisioning and/or reconfiguration to improve the VoIP quality. In this work, we use SAVI SDI to prototype a novel method of enhancing VoIP quality by using its converged and centralized manner of resource management. We will demonstrate that with a reconfiguration operation from the SDI platform, a bandwidth thirsty HD video call can be elevated to a higher QoS in real time to achieve visible significant quality improvement compared to a call made over best effort VoIP connection.

An Orchestration Service of SAVI Testbed; A Heat Approach

Honbin Lu, Hadi Bannazadeh, Alberto Leon-Garcia, University of Toronto

Heat is an OpenStack project for providing an orchestration service to facilitate the provisioning and management of OpenStack’s resources. A key feature of Heat is to provide a domain specific language (DSL) for users to declare the list of cloud resources that they need. By integrating Heat to SAVI Testbed, users are provided a tool to automate the resource provisioning process, and manage the complexities of application deployment and management at Testbed.

Big Data as a Service for SAVI Testbed

Honbin Lu, Hadi Bannazadeh, Alberto Leon-Garcia, University of Toronto

Hadoop is a popular software framework that supports a simple programming model for Big Data processing. Sahara is a service for supporting Hadoop at OpenStack clouds. Sahara reduces the complexities of provisioning and managing Hadoop clusters by implementing the common provisioning and management routines. By integrating Sahara into SAVI Testbed, users are provided a tool which facilitates the provisioning of Hadoop clusters, and allows user to focus on the high-level specification of their Hadoop clusters without distracting from the low-level workflow and configurations.

Cloud-RAN on SAVI; a GSM approach

Mohammad-Sina Tavoosi-Monfared, Hadi Bannazadeh, Alberto Leon-Garcia, University of Toronto

This poster accompanied by a demo presents an implementation of OpenBTS (Open Base Transceiver Station), a software defined GSM on SAVI Testbed. The equipment used include an 8-channel OctoClock clock distribution module, four USRP N210 and an integrated GSPDO kit. A phone conversation with two cellular phones will be tested. To simulate a C-RAN (Cloud-RAN) architecture, the transceiver module will be implemented on a closer testbed edge while the OpenBTS module (i.e. routing) will be on a further module that is not delay sensitive.

Monitoring and Measurement as a Service in SDI deployed on SAVI testbed

Jie Yu (Eric) Lin, Rajsimman Ravichandiran, Hadi Bannazadeh, Alberto Leon-Garcia, University of Toronto

IT infrastructure monitoring and measurement is an important aspect for Intelligent management and autonomous infrastructure. In this demo, we will demonstrate the monitoring and measurement (M&M) system in SAVI testbed. Following the Software Defined Infrastructure (SDI) concept, this system monitors both compute resources and network resources (including virtual and physical resources). It also provides mechanism for monitoring user specified metrics (e.g. web server status). The system takes care of collecting, processing and storing all of the monitoring data including the historical records. It provides both the admin and the user visibility to the status of the infrastructure.

Efficient Multicast algorithm on SAVI network

Sai Qian Zhang, Hadi Bannazadeh, Alberto Leon-Garcia, University of Toronto

Group communication is widely applied in the daily computer network system. When more than one receiver is considered in the data transmission mechanism, multicast is the most efficient and reliable solution. In this poster, we proposed an efficient multicast algorithm used among different virtual machines (VMs) in the SAVI network. The algorithm best fits for the dynamic network topology so that when new VM joins the multicast network and the topology changes, we do not need to reconfigure the whole multicast tree topology. In particular, the multicast tree is constructed based on the following objectives: 1) Saving Bandwidth 2) Minimizing the delay.

End-to-End Traffic Control in the SAVI Testbed

Thomas Lin, Hadi Bannazadeh, Alberto Leon-Garcia, University of Toronto

This demo will show an example of how the SAVI testbed enables end-to-end traffic control as managed by the user. For the purposes of this demo, we have integrated a Wi-Fi access point (WAP) that also allows mobile devices to be integrated into a virtual network. The SAVI testbed currently enables users to control their own virtual networks using an OpenFlow controller running with a publicly accessible IP address. The demo will show-case a scenario in which the mobile device is connected to the testbed and receives a live video stream from a virtual server running within the testbed, while the network is being congested. Via their own OpenFlow controller, the user is able to direct the traffic through a pre-configured queue that guarantees a certain level of available bandwidth. The demonstration will aim to show the effect of queuing on the quality of the displayed video on the mobile device.

L2/L3 Overlay Software Defined Network on SAVI Testbed

Khashayar Hoseinzadeh, Hadi Bannazadeh, Alberto Leon-Garcia, University of Toronto

In this work, we will utilize the SAVI testbed alongside a software defined switch (OpenVSwitch) to dynamically create network topologies specified by the user. Similar to mininet, the users will be able to specify the number of hosts and switches and specify how they want them to be connected. The goal is to create a layer 2 or a layer 3 network and linking up the nodes as defined by the user topology. For this purpose, we use SAVI TB and VXLAN tunneling protocol to create arbitrary overlay networks. These overlay networks could be used in variety of applications from deploying a large scale OpenFlow network in order to test a SDN controller and application to running a large scale distributed application on a desired topology.