1. Network Resource Broker for IPTV in Cloud Computing Lei Liang, Dan He University of Surrey, UK D.He@surrey.ac.uk OGF 27, G2C Workshop 15 Oct 2009 Banff, Canada

2. Outline Introduction Network Resource Broker Testbed Scenario and Validation Conclusion and Future Work

3. Introduction The evolution from Grid Computing to Cloud computing has introduced a very important add-on value service that is on- demand resource provisioning The Cloud computing is to use data centre that surely requires QoS provision to network services. The valuable functions of Cloud may bring technical challenges to the existing networks: Network extensibility and scalability due to service composition How will it impact the network in terms of customibility Can we reuse some of the existing Grid-Network enhancements Challenges and opportunities on network reliability and security

4. EC-GIN Project EC-GIN project was founded by the EU committee to tackle some of these questions with focus on the impact to the existing networks. Its objective is, based on a number of properties that make Grids unique from the network perspective, to aims at developing a tailored network technology in dedicated support of Grid applications. It is to supplement with a secure and incentive-based Grid Services network traffic management system, which will balance the conflicting performance demand and the economic use of resources in the network and within the Grid Although the research is carried out for Grid computing, we found it also applies to the Cloud computing with helps to provide customized QoS in networks with efficient network resource utility.

5. Resource Broker for Cloud Although the research is carried out for Grid computing, we found it also applies to the Cloud computing with helps to provide customized QoS in networks with efficient network resource utility. The resource management system plays a critical function in cloud in terms resources discovery for users and allocations of the best resources to the users tasks Many researches efforts have been put on the resource broker, each of which has targeted applications resource allocations for different applications have various criteria No general approach Network resource broker proposed here is for data shifting applications

6. Network Resource Broker The Network resource broker is designed For data shifting applications To provide optimized date transmission path in networks To use network bandwidth efficiently, and to satisfy uses timing requests (QoS) The application could be pure file transmission or real time multimedia streaming The main functions of the resource broker includes: Interpreting the requirements of users Obtaining correct resource availability information Acquire and process network topology information Calculate optimized paths for data transmission using overlay routing concept for bandwidth utility efficiency and task timing requests Coordinate with bandwidth reservation function unit for bandwidth reservation on the optimized paths Coordinate with data transfer function unit for data transmission

7. Broker Architecture The resource broker has 6 interfaces: Two interfaces between users and the broker for inputting users jobs and outputting job execution confirmation One interface between the resource brokers data resource discovery function and the data resource index service unit for data resource availability and location enquiry and feedback One interface connects the brokers network interpreter function and the network topology matrix which provide bandwidth information for every link between each pair of hosts One interface between the transmission path optimizer and the bandwidth reservation unit reserved bandwidth on the optimized paths The last interface connects the Transmission Scheduler and Executer function and the file transfer function unit

8. Testbed The proposed design has been implemented on a testbed The testbed is formed by 7 Linux boxes with Ubundu 8.10 installed 3 routers connecting 4 end users Router 1, 2 and 3 are connected with each other by 100Mb Ethernet interfaces forming a triangle End users Host B connected to Router 2, Host A to Router 1 and Host C and D both to Router 3 by 100 Mb Ethernet interfaces. every end user is in a different sub-net The designed network resource broker is installed in Host B

9. Testbed layout Host A Host D Host B Host C Router 1Router 3 Router 2 Video_S Video_B Default and optimized path for Video_S Optimized path for Video_B Default path for Video_B

10. Network resource broker in IPTV Scenario The test scenario is designed for potential user cases for IPTV big streaming file splitting for multiple path transmission, and the current simplified case for content synchronization An IPTV provider have several content servers located in different places to cover large area for better streaming quality on the Internet, which are need be synchronized a specified time with limit length so the impact to the service can be minimized, such as the early morning every day. Network resource broker can be used for fast and efficient content shifting among servers

11. Validation Host B is acting as one of the server that is trying to synchronize itself with other servers Host C and Host D A large video file from D A smaller video file from C Done in the same time Link between Router 3 and Router 2 doesnt have enough bandwidth for both the transmission in time. The bandwidth between Router 3 and 1 and between Router 1 and 2 are even smaller that can not finish the downloading of the large video in time The user request is processed by the network resource broker that Queried the resource index file to discover that Video_B is on Host D and Video_S is on Host C process the path on which these two files should be downloaded that can satisfy the timing requested by the user using the network topology matrix as an input

12. Results a) Optimized path for large video file b) Optimized path for small video file Results shows that the large video file should go through the link between Router 2 and Router 3, and The small video file should go through the link between Router 3 and Router 1 to Host A first and then be forwarded by Host A to Host B through the link between Router 1 and 2 The path optimization satisfied the users timing request by using overlay routing through Host A The result showed that both files were successfully received by Host B on time with variation of several ms

13. Conclusion and Future Work A network resource broker as one of the central functions of the resource management system for Cloud computing is presented It can choose data transmission path for Cloud computing tasks in the network to satisfy the users timing requests with QoS guarantee while using the network bandwidth more efficiently by applying overlay routing concept to the path processing The proposed resource broker has been implemented in Java and validated in an IPTV scenario on a testbed It successfully scheduled two video files transmissions and optimized the downloading paths so the bandwidth bottleneck is avoided and two transmissions were finished on the same time as requested by a user We are trying to use the Large File Transfer (LFT) mechanism developed in the EC-GIN project to replace netcat as the data shifting tool the peer awareness mechanism developed in the same project will be used to provide dynamic network topology matrix as an alternative for the static approach we are currently using

14. Thanks