2016 International Conference on Electronic Information Technology and Intellectualization (ICEITI 2016) ISBN: 978-1-60595-364-9 A Strategy of Traffic Optimization Based on the Technology of SDN Yirong Wang, Hongkai Wang, Botao Yu and Yue Ma ABSTRACT Modern Content Delivery Network()[1,2] by caching and scheduling strategy to reduce user-perceived latency, because can schedule user to the cache node which closest to. At the same time, this strategy can reduce the data flow which has to be sent back to the source station, and decrease the calculation pressure of source station. For some content which is dynamic and cannot be cached, the current acceleration approach is to create a GRE tunnel between the nodes, using the static route scheduling dynamic data flow, improve the speed of data flow which back to the source station. It has been a recent trend of to separate the control flow and data flow through technical, in order to change traditional dynamic/static content acceleration model, to optimize the traffic. The core idea of Software Defined Network(SDN)[3,4]is through the network control plane and data plane separation equipment, so as to realize the flexible control of network traffic. SDN technology is used to transform the, Make the have an intelligent traffic scheduling environment which can focus on monitoring network and scheduling flow path real-time According to the different business. Ultimate aim is to avoid local congestion, optimize the traffic, and improve the quality of the service effect. Yirong Wang, Botao Yu, Yue Ma, Beijing Guodiantong Network Technical Co. Ltd., Beijing China Hongkai Wang, Information and Communications Branch, State Grid Zhejiang Electric Power Company, Hangzhou, China 211
INTRODUCTION By adding a layer in the existing Internet intelligent virtual network, pushing the contents to the virtual network node which is most close to the user, allows a user to obtain the content from the nearest node, so as to solve the congestion problem of internet network, and improve the response speed of the user to access the website, make the IP network into a high efficient, reliable, intelligent network from which was disordered, inefficient and unreliable originally. In the SDN architecture, the control and data planes are decoupled, network intel ligence and state are logically centralized, and the underlying network infrastructure is abstracted from the applications. Traditional Internet architecture is distributed, each network device has a relatively independent of the operating system and control level, exchange information and data between equipment by distributed Internet protocol. After adopting SDN architecture, we can do the centralized scheduling based on global network resources. Under the environment of SDN, traffic engineering[5] can bring exciting increases for network performance, the existing successful cases, such as Google's B4 system [7]has used dedicated SDN equipment to trans formed the network of data center into SDN completely, this project made the network utilization be promoted to nearly 100%. After using general SDN equipment, the software definition WAN (Soft WAN, SWAN) system [6] of Microsoft made the network utilization be increased by 60% ~ 70%. STRATEGY OF TRAFFIC OPTIMIZATION The Cache Cluster Collaborative Interaction Optimization For getting the information of contents which were cached in nodes, it is necessary to establish a good communication channel between the vast nodes after constructed to a certain size. Copy content to cache from neighbor node more efficient rather than from the source station, and can save the data traffic for source station. We can improvement for Cache cluster collaborative interaction base on network architecture of SDN: 1.Develop schedule APP by SDN north interface, all cache information of nodes are stored in the schedule APP. 2. schedule APP can get global traffic information of from SDN controller in real time. Figure 1 shows the process of Cache cluster collaborative interaction base on SDN: 212
Schedule APP 3) 4) SDN Controller 2) 5) Source Station SDN Switch Interaction 1) Data transmision 6) 7) node1 node1 Figure 1. Cache cluster collaborative interaction base on SDN. 1) node2 received the content request packets come from user, because the content does not exist in local cache, node 2 have to simulate client to send content request. 2) the content request packets come from node2 arriving at SDN switch the node 2 access to, the switch found that there is no matching flow table item after lookup flow table, and then encapsulate request packet into Packet in message up to SDN controller. 3) SDN controller judge the type of packets after receiving the message in order to send these packets to the corresponding schedule APP. 4) schedule APP judge the source IP of these packets is belong to node2, and lookup the cache information summary for finding out the node which have cached the contents node2 need. If there are multiple results can be matched in summary, schedule APP will provide the node which have best network condition in this moment from the matching results, such as node 1. 5) schedule APP send Flow-mod message to SDN switch. 6) Switch add flow table item after received the message, and redirect the request packets come fromnode2 to the node1 through the flow table item. 7) Node2 pull content and cache it in local from node1. 213
Traffic Scheduling Based on the Quality of Nodes Bandwidth This paper presented a traffic scheduling optimization strategy based on the node of the quality of the bandwidth. Path quality detection can obtain more links indicators, such as delay, jitter, packet loss rate, etc. The process is as follows: 1.At the beginning of the construct 1) SDN Controller obtain the topology information of global network, connected network forward equipment by automatic or manual. 2) Path quality detection detect link quality in turn by traversing all path in. 3) Path quality detection report the detection data to APP. 2.User requests resources 1) After receiving the message of user request resource SDN controller send the packet to corresponding scheduler APP by the type of packet. 2) APP match the link quality table, calculate the optimal business path between user and a cache node, then the bandwidth and priority information be unified recorded in user forwarding table as forwarding strategy. 3) APP send Flow-mod message to SDN switch. 3.In the process of running 1) APP monitors the quality of all links in the network, and update the entire network link - quality table. 2) Forward equipment reports flow data based on user information or user business to SDN controller according to a specified period. 3) APP records user business traffic in statistics table. 4.Whenquality degradation appeared on a single link 1)The weighted score of delay, jitter and packet loss rate is use for judgment method of deterioration in the quality. 2) APP generated the scheduling strategy according to the user's priority and business priority, and first we performed user's priority. 3) APP sent the computation to SDN controller, SDN controller generate forwarding flow table and sent it to SDN switch. Dynamic Content Acceleration Based on The Technology of SDN This paper proposed a dynamic content acceleration strategy based on the technology of SDN. Figure 2 shows the create process of SDN acceleration tunnel: 214
Bandwidth priority weighting Flow group information Bandwidth allocation function Dynamic flow Back to source station abstract Back to the source tunnel optimization node topology information Topology information gathering Tunnel and tunnel group information Tunnel database management Traffic scheduling operation Topology and link bandwidth information SDN Gateway Node port state change Traffic scheduling operation User characteristic information back to source station Dynamic flow back to source station Figure 2. The create process of SDN acceleration tunnel. 1. Dynamic flow back to source station abstract sent characteristic information of the flow to bandwidth priority weighting. 2. Input the result to the called back to the source tunnel optimization. 3. At the same time, back to the source tunnel optimization received the node topology information reported by SDN gateway after topology information gathering extracted, the node topology information is another input for optimization algorithm. 4. Finally the back to the source channel optimization according to the result of TE optimization algorithm to decide the back resource station flow will use which tunnel and tunnel group. 5. Tunnel and tunnel group information will continue to be input into the tunnel database management, tunnel database management and out the flow schedule operators through SDN gateway. 6. TE optimization algorithm output tunnel group information for different business, control the business flow with different outer tunnel encapsulation, so as to control the bandwidth allocation of dynamic flow. RELATED WORKS There have been many works that attempt to optimize the traffic: The Cache cluster collaborative interaction optimization: Loosely coupled cache communication protocol which based on the network messages is used in 215
collaborative interaction between nodes, such as IC, HTCP, Cache Digest[8],Cache Pre-filling [9]and so on. Traffic Scheduling Based on the Quality of Nodes Bandwidth: Traffic engineering based on SDN is the cutting-edge technology in researches of traffic scheduling based on bandwidth quality in the present. SDN is applied on telecom operator s backbone traffic scheduling is inspired by the Google B4 project at the earliest. Google have long noticed the value of SDN technology for the Internet data center, as early as in 2009, Google vigorously working in SDN/Open Flow exploration/practice just contact SDN/Open Flow concept. The based on the SDN technology: Researchers put forward ALTO agreement in 2011, ALTO provide the interface for changing network status, realize the flow optimization in SDN architecture according to this API. CONCLUSION Reference SDN successful application in traffic engineering technology, optimize the flow, through the centralized scheduling function of SDN controller, realized the link quality and bandwidth utilization detected and collected automatically, we can created GRE tunnel automation on-demand, and create the forwarding path automatically. Develop scheduling APP to realize global flow schedule function according to the north interface by SDN controller providing, schedule strategy is constructed by the detection results of path quality detection and the information of the priority of the data traffic; dynamic content accelerate based on SDN realized manager route and tunnel automatically, and can choose the optimum route in the situation of multipath, reduce the effects of the default routing congestion delay when busy. REFERENCES 1. networks Inc. http://cdnetworks.com/. 2. Content Delivery Networks Interconnection.http://tools.ietf.org/wg/cdni/. 3. Weifeng Zhang, 2014, Depth Resolution of SDN, Beijing: Publishing House of Electronics Industry, pp. 35-42. 4. Jain S., Kumar A., Mandal S., 2013, B4: Experience with a globally-deployed software defined WAN, ACM SIGCOMM Computer Communication Review, 43(4): 3-14. 5. Aw Duche D., Chiu A., Elwalid A., 2002, Overview and principles of Internet traffic engineering, RFC3272, pp, 28-36. 6. Hong C.Y., Kandula S., Mahajan R., 2013, Achieving high utilization with software-driven WAN, ACM-SIGCOMM Computer Communication-Review, 43(4): 15-26. 7. Danana E., Hassidim A., Kaplan H., Kumar A., Mansour, Y. Raz D. and Segalov M., 2012, Upward Max Min Fairness, INFOCOM, March, 2012: 837-845. 8. http://www.squid-cache.org/cachedigest/cache-digest-v5.txt. 9. http://www.cs.technion.ac.il/~rcohen/papers/cache-pre-filling.pdf. 216