request routing in cdn
TRANSCRIPT
by:
Sandeep Kath
Improving the Request Routing Mechanism in Content Delivery Networks Improving the Request Routing Mechanism in Content Delivery Networks (CDNs)(CDNs)
Outline Content Delivery – a bit of History Content Delivery Network Request Routing Techniques Literature Review Problem Formulation Research Objectives Methodology System Description Performance Metrics Results and Discussion Summary of the results Recommendations Conclusion Future scope of the work Publications References
Content Delivery – a bit of history
• Individual Web servers• Increase in Web content
• Web Server Farms• Issue of Flash Crowds
• WWW has evolved beyond simply displaying static webpages
• New challenge of delivering the content like realtime audio, video etc.
Individual Web Sever (Non-CDN)
Content Delivery Network (CDN)
• What: Geographically distributed network of Web servers around the globe.
• Why: Improve the performance and scalability of content retrieval.
Content Delivery Network
Technology Components
• Content distribution– Placing the content to the devices
• Request routing– Steer users to a delivery node that is close
• Content delivery– Protocol processing, access control, QoS mechanisms
• Resource accounting– Logging and billing
Request Routing(An Example)
Request Routing Techniques
• Global Server Load Balancing (GSLB) • DNS-based request-routing• Transport Layer Request Routing • HTTP redirection• URL rewriting • Any casting • CDN peering
DNS –Normal Operation
DNS Request Routing
Problem Formulation• DNS based request routing does not use IP address of client.
The knowledge of Internet location of the client limits the ability of the request-routing system to determine a client’s proximity to the surrogate.DNS query does not carry the addresses of the querying client
• DNS server redirects the resolved IP address to the client and client uses that IP address to connect with the server, so this current flow increases network latency.
• The main disadvantage of DNS-based request-routing is that, it increases network latency because of increase in DNS lookup times.
Objectives
1. Comparison and performance evaluation of CDN network with non-CDN network on basis of three metrics Request return time (RTT), Packet Loss, throughput.
2. Analyze the various request routing techniques for Content delivery networks and detailed study of DNS based routing and transport layer routing techniques.
Objectives…
3. Design new request routing technique which will have less Request response time, eventually this new technique will have less packet loss and throughput.
4. Performance comparison of new proposed request routing with CDN and non-CDN network.
5. Implement the local load balancing on servers by using least connections technique.
Methodology Used
• Study and review of DNS and name resolving techniques.
• Detailed Analysis of request routing techniques.• Exploring the various load balancing techniques.• Work with NS 2 and learn TCL and NMAP.• Design of new protocol or agents.• Implementation of new agents in NS2 /C++.• Implementation of Simulation scripts in TCL/TK.• Implement AWK scripts to get results from the NS 2
trace files.• Results comparisons.
System Description
System Description– Bandwidth at each link
Link Between Bandwidth Delay
DNS Server & Content
Server
3 Mbps 50ms
DNS Server & Client 1 Mbps 50ms
Hop & DNS Server 6 Mbps 50ms
Hop & Router 1 Mbps 50ms
Router & content server 3 Mbps 50ms
System DescriptionFive New CDN Agents and packets developed in NS-2 to simulate CDN:
S.NO. AGENT PACKET TYPE
1. cdn_client PT_CDN_CLIENT
2. cdn_dns PT_CDN_DNS
3. cdn_server PT_DNS_SERVER
4. cdn_router PT_CDN_ROUTER
5. non_cdn_dns PT_NON_CDN_DNS
6. cdn_dns_new PT_CDN_DNS_NEW
System DescriptionProposed CDN DNS Request Routing Technique
System DescriptionHash Table
Source Address Content Type Destination Address
1 203.197.219.7 txt/plain 202.54.122.1
2 103.194.220.4 txt/plain 202.52.123.2
3 202.54.103.1 txt/html 202.54.122.1
…
n Source_Addr Content_Type Destination_Addr
Hash Functionindex = ((source – 10) * 10 + content_type ) % length of table
System Description – CDN DNScdn_dns_new agent has been developed and added source, content_type and object_id as input parameters. The source field represented by nsaddr_t (address type used on NS-2), defines the IP address of the web server found to serve the client. Content_type represented by an integer, defines the type of content requested by cdn_client. Object_id represented by an integer, defines the size of content requested by cdn_client.
System Description – Local Load Balancing in CDN
Round Robin Round Robin algorithm distributes requests to the different web servers in a round robin approach.
Least Connectionsit keeps track of the number of active connections to each server and always directs a new connection to the server with least connections.
Round TripIt monitors the request/respnose phase of each connection by monitoring the TCP protocol. For each connection, the elapsed time between forwarding the first byte of the request to the server and first byte of the response to the client is calculated.
System Description – Local Load Balancing in CDN
System Description – Local Load Balancing in CDN
Performance Metrics • Round-Trip Time
• Round-trip time (RTT) is the total time taken for a packet sent by a node A to reach a destination B and then for a response to sent back by B to reach A. In other words, the round-trip time is the sum of the one-way delays from A to B and from B to A, and of the time it takes B to formulate the response to the original packet. Round-Trip Time represents response time and related to delay.
Performance Metrics • Packet Loss
• Packet loss is determined as the probability of a packet being lost in transit from a source A to a destination B.
• Main reasons of Packet Loss– Congestion– Errors
• In multimedia applications packet loss is tolerated but in TCP, packets are retransmitted if they are lost and thus reduce the performance
Performance Metrics • Throughput
– Throughput is the average rate of successful message delivery over a communication channel. The throughput is usually measured in bits per second (bit/s or bps), and sometimes in data packets per second or data packets per time slot.
– The system throughput or aggregate throughput is the sum of the data rates that are delivered to all terminals in a network.
Performance Metrics • Server Load
– Load expresses how many processes are waiting in the queue to access the computer processor. This is calculated for a certain period of time and of course, the smaller the number, the better.
Performance Metrics • Hop-Count
• The number of point-to-point links in a transmission path.
• Each point-to-point link is technically a hop, the hop count is the number of network devices between the starting node and the destination node.
• An IP packet traveling over the Internet can easily "hop" through more than a dozen routers.
Results- CDN vs Non-CDN
Packet Loss
Analysis – Content Delivery Network
• There are more packets lost in the Non-CDN network than in the CDN network.
• In CDN network, loss of packets only occurred when clients’ requested contents are not stored in the nearest content servers.
Results- CDN vs Non-CDN
Throughput
Analysis – Content Delivery Network
• Throughput of the link between the hop (node n1) and DNS server (node n3). The upper curve shows throughput in Non-CDN network, the lower curve shows that in CDN network.
• The throughput in Non-CDN network is higher than that in CDN network, showing that the link in Non-CDN network is busier.There are more packets lost in the Non-CDN network than in the CDN network.
Results- CDN vs Non-CDN
RTT
Analysis – Content Delivery Network
• At the beginning of the curve Non-CDN and CDN behave the same since the clients’ requests are served by servers at the far end, and that for the rest of the curve Non CDN has a larger RTT than CDN since clients in CDN can be served by servers nearby.
Results- Proposed CDN DNS Request Routing
Packet Loss
Analysis – Proposed CDN-DNS Request Routing
The upper curve shows packet loss in Non-CDN DNS network, middle curve shows the packet loss in normal CDN with existing DNS and the lower curve shows the data in CDN DNS network.
Results- Proposed CDN DNS Request Routing
Throughput
Results- Proposed CDN DNS Request Routing
Round Trip Time
Results- Local Load Balancing in CDN2-Node Cluster 4-Node Cluster
8-Node Cluster
Results- Local Load Balancing in CDN
Round Robin, Round Trip and Least connection techniques on different clusters
Analysis – Local Load Balancing
• Figure illustrates that the round trip scheme outperforms the other schemes for simulated CDN network.
• From the comparison it is clear that the round-trip scheduling strategy gives the lowest server load and lowest average response time followed by the round robin strategy
Summary of Results 1. Our comparison scenarios have illustrated that CDN performs
better than Non-CDN in terms of packet loss, RTT and throughput.
2. CDN-DNS request routing is more efficient as compared to existing DNS redirection. It uses the client’s location to redirect the request to appropriate server. Proposed DNS request routing technique saves one cycle of DNS resolution, in which instead sending the resolved IP address to client, client’s request is directed to surrogate server of CDN along with client information and surrogate serves the requested content to the client
Summary of Results 3. Also we have analyzed that low bandwidth links in the network could cause
more packet loss in existing DNS request-routing than proposed CDN DNS technique as traffic in proposed scheme CDN is to be distributed.
4. The round trip scheme outperforms the other schemes for local load balancing in simulated CDN network. From the comparison, it is clear that the round-trip scheduling strategy gives the lowest server load and lowest average response time followed by the round robin strategy.
Recommendations • Simulation results reveals that CDNs can provide the content to
many users efficiently and reliably even at times of maximum Internet traffic or during flash crowd.
• Simulation results indicate that CDNs optimize the speed or increase the response time by keeping the content near to user location.
Recommendations • Simulation results indicate that CDNs can provide fail-safe feature
by keeping the same content on multiple servers or content redundancy.
• Simulation results indicate that present DNS request routing have limitations and it can not use user location to resolve the domain name so that traffic should be directed to nearby server. Results also reveal that proposed CDN-DNS is more efficient as compared to existing DNS request routing.
• Simulation results indicate that Round Trip load balancing technique performed better in case of local load balancing in CDNs.
Conclusion A new CDN DNS technique, proposes new DNS
routing table to resolve client requests as per client location and content type. This CDN DNS request routing technique can efficiently route request of the client to nearby CDN server, which has less round trip time (RTT), less packet loss and less throughput. It can greatly reduce time delay of the request routing, thus effectively ease Internet congestion. This technique is helpful in researching more efficient DNS systems & caching for CDNs and has better application foreground.
Future ScopeIn this thesis, we have considered simple hashing algorithm to resolve the request and this work can be extended by applying to more efficient hashing algorithms.
Investigations can be extended by taking different types traffic like IPTV, Video on Demand and real audio.
In this thesis, single DNS server has been used to carry out the research work, we can extend and do the performance analysis of CDN network if we have multiple DNSs on same network and how these DNS server will interact with eachother to resolve the address.
In this thesis, we have analyzed the request routing techniques for CDNs, Content replication and caching are another major components of a CDN. This work can be extended to simulate and analyze the content replication & caching techniques on CDN.
In this thesis, we have considered Round Robin, Round Trip and Least connections load balancing techniques. This work can be extended by applying dynamic and adaptive load balancing techniques.
Publications
• Sandeep Kath, Manoj Kumar and Ajay Sharma, “CDN DNS - An Efficient DNS Request Routing Technique in Content Delivery Networks” International Journal in Advances in Computational Sciences and Technology , ISSN 0973-6107 Vol 3 Number 2 (2010) pp. 147-154.
References• Al-Mukaddim Khan Pathan and Raj Kumar Buyya , “A Taxonomy and survey of content
delivery networks”, in Content Delivery Networks, Chapter 2, Springer Press, ISBN 978-3-540-77886-8.
• Rajkumar Buyya et al., “A case of Peering of Content Delivery Networks”, Distributed Systems Online, IEEE, vol. 7, issue 10, pp. 3-3, Nov 2006.
• Chengdu Huang and Abdelzaher T., “Towards content distribution networks with latency guarantees”, Quality of Service, 2004, IWQOS 2004, 12th IEEE Workshop, vol. 7, issue 9, pp. 181–192, June 2004.
• G. Peng, CDN: Content Distribution Network, Technical Report TR-125, Experimental Computer Systems Lab, Department of Computer Science, State University of New York, Stony Brook, NY 2003.
• Ao-Jan Su et al., “Drafting Behind Akamai”, ACM SIGCOMM Computer Communication Review, vol. 36, issue 4, pp. 435-446, 2006.
• Akamai. Akamai content delivery network. http://www.akamai.com.• C. Desem and P. L. Chu, “The effectiveness of request redirection on CDN robustness”, in
proceedings of the 5th symposium on Operating systems design and implementation OSDI ’02, vol. 36, issue SI, pp. 345-360, 2002.
• Kirk L. Johnson , “The measured performance of content distribution networks”, in Computer Communications (COMCOM) Journal, vol 24, issue 2, pp. 202-206, Feb. 2001.
• Limin Wang et al. , “Reliability and securing in the CoDeenN Content Distribution Network", in proceedings of the annual conference on USENIX Annual Technical Conference, pp. 14-14, 2004.
References• Yun Bai et al., "An Efficient Load Balancing Technology in CDN", in International Journal of
Applied Mathematics and Informatics, vol. 1, issue 2, pp. 92–96, 2007.• Mohit Aron et al., "Scalable content aware request distribution in cluster based networks
servers", in proceedings of the annual conference on USENIX Annual Technical Conference, pp. 26-26, 2000.
• Tsang-Long Pao and Jian-Bo Chen, “Capacity Based Load Balancing Scheme for Fair Request Dispatching”, in Asian Journal of Information Technology, vol 5, issue 11, pp. 1284-1290, 2006.
• Zhao W. and Schulzrinne H. ,”DotSlash : Handling Web Hotspots at Dynamic Content Websites ”, in 25th INFOCOMM - IEEE International Conference on computer communications, vol. 23, issue 29, pp 1-5, April 2006.
• Hoakan Bryhni et al. , “A Comparison of Load balancing techniques for scalable web servers”, Network IEEE , vol. 14, issue 4, pp. 58-64, July-Aug 2000 .
• Adrian Popescu et al., “Content distribution Over IP : Developments and Challenges”, in Joint EuroNGI anf ITG Workshop on Visions of Future Generation Networks, 2006.
• Van Rompaey et al. , “Bandwidth versus storage trade-off in a content distribution network and a single server system” in 7th conference ConTEL 2003, Telecommunications , vol. 1, pp. 315-320 , 2003.
• Cardellini et al. , “Geographic load balancing for scalable distributed Web systems” in 8th International Symposium on Modeling, Analysis and Simulation of Computer and Telecommunication System, pp. 20-27, 2000.
• Valeria Cardellini et al. , “Dynamic Load Balancing on Web-Server Systems ", IEEE Internet Computing, vol. 3, issue 3, pp. 28-39, 1999.
References• Colajanni M. et al., “Dynamic load balancing in geographically distributed heterogeneous Web
servers ”, in Proceedings of 18th International Conference on Distributed Computing Systems, vol. 26, issue 29, pp. 295-302, 1998.
• Paul V. Mockapetris and Kevin J. Dunlap “Development of the Domain Name System”, in the proceedings of SIGCOMM ’88 Computer Communication Review, vol. 18, issue 4, pp. 123-133, August 1988.
• S. Manikamdan et al. , “A Framework for minimizing latency in CDN using URL request routing approach”, in Academic Open Internet Journal , vol 16, 2005.
• Balachandar Krishnamurthy, Craig Wills and Yin Zhang , "On the Use and Peformance of Content Distribution Networks", ACM SIGCOMM Internet Measurement Workshop, 2001
• Cherkasova et al., "Performance analysis of content-aware load balancing strategy FLEX: two case studies ", in 34th Annual Hawaii International Conference on System Sciences, vol. 9, 2001.
• Zhougju Zhang and Weiguo Fan , “Web server load balancing : A queueing analysis ”, Elsevier – European Journal of Operational Research, vol. 186, pp. 681-693,2008.
• Zhuoging Morley Mao et al. , “A Precise and Efficient Evaluation of the Proximity between Web Clients and their Local DNS Servers”, USENIX Annual Techincal Conference, pp. 229-242, 2002.
• Valeria Cardellini et al., “Redirection Algorithms for Load Sharing in Distributed Web-server Systems”, 19th IEEE International Conference on Distributed Computing Systems, pp. 528-535, 1999.
• Mukaddim Pathan et al. , “Load and Proximity Aware Request-Redirection for Dynamic Load Distribution in Peering CDNs”, in Proceedings of the OTM 2008 Confederated International Conferences, CoopIS, DOA, GADA, IS, and ODBASE 2008. Part I on On the Move to Meaningful Internet Systems, vol. 5331 (2008), pp. 62-81, 2008.
Thank YouThank You