NDLB: Nearest Dispatcher Load Balancing approach for Web Server Cluster

Author Name(s): Kadiyala Ramana, M.Ponnavaikko, A.Subramanyam
Author Email: ramana.it01@gmail.com


With the growing popularity of web based applications, the primary and consistent resource in the infrastructure of World Wide Web are web server clusters. Overtly in dynamic contents and database driven applications, especially at heavy load circumstances, the performance handling of clusters is a solemn task. A novel distributed web server system NDLB (Nearest Dispatcher Load Balancing) is proposed in this paper which uses both DNS and Dispatcher to forward the client requests efficiently to the servers in a user transparent way. This system conquers superior response time than other distributed web server architectures and also poises loads between servers within the clusters effectively. However, the NDLB architecture is accessible and more indulgence in both Dispatcher and DNS; Moreover, if the cluster capability is less than the request rate it offers a load balancing architecture.


The volume of the information available online and services available for the internet users increased with the blast of the world wide web. The thriving of information and various service demands has made a sensational pressure on the World Wide Web (WWW) infrastructure. To serve a large number of client request they need advanced web server systems. Because of their scalability, availability and cost-effectiveness distributed web server cluster architectures became more popular instead of using one web server, which has high processing capabilities.

In 1995, the number of internet users was less than 1% in the world population, whereas today it is 40%. In 2016, there were 3.5 billion internet users while in 2005 there were 1.02 billion internet users [1]. With the fast growth of internet traffic, most popular websites need to scale up their server capacities. The popular way to provide a list of alternative, or equivalent mirrored servers at different locations. The mirrored servers are not transparent to the users and it is hard to provide load balancing and fault-tolerance [2]. The technique which is used to redistribute the workload from loaded servers to idle servers in order to improve the performance is called Load balancing. The most promising approach to handle popular web sites is to maintain a virtual single interface and to use a distributed architecture. A web cluster is known to be a compilation of servers which works jointly as a solitary articulate system for providing highly & scalable web services. It relies on load balancing techniques where it shares service traffic efficiently between its back-end servers and visibly to the clients. The scalability is termed as the capacity in system measurement where to meet the escalating demands as service traffic. The capacity of the system is determined based on the support of number of parallel connections of servers per second without affecting of momentous queuing delay in the interior infrastructure.

By taking advantage of the server redundancy, load balancing techniques improves the system availability [3]. The ability of a server to provide endless services over time is called Availability and it is deliberated as uptime percentage.  When a cluster server declines or abort, the load will routinely redistribute with slight or refusal brunt laying the service among other available services. The servers in the Web server cluster are not essentially situated in the equivalent site and they will be located in diverse biological locations. In proxy servers they are all located at different locations. Because of the rapid increase of Internet, the broadcast time is an important recital factor in network service. In web cluster, load balancing involves a several major concerns. The primary concern is measurement of work load. In different applications, workload has different meanings. In web services, the client request is a basic building block of load balancing and its response lively connections is a simple server load index [4]. Request distribution policy and mechanism are the two additional core issues in the load. For each incoming request from the clients, the load balancing policy will determine the target server allocation policy competently and evidently to clients. Numerous load balancing mechanisms and policies are available with diverse characteristics.


In this paper a novel approach is proposed for dynamic load balancing with both DNS and Dispatcher. DNS calculates the round-trip time to the dispatcher of each cluster and forwards the IP address of the Dispatcher, to the client which has the low round trip time. Dispatcher selects the appropriate server in the cluster using random method. A model web server cluster was employed and equipped with the proposed algorithm. The investigational results attained from the JMeter tool confirm the enhancements in clusters performance in terms of CPU utilization, Error Rate, Average Response Time, Number of Requests served and Throughput in contrast to the DNS, Dispatcher and Dynamic Dispatcher based approaches. This approach also provides availability and scalability when compared to the existing approaches.


  1. http://www.internetlivestats.com/internet-users/
  2. Kadiyala Ramana and M.Ponnavaikko, “Web Cluster Load Balancing Techniques: A Survey”, International Journal of Applied Engineering Research, Volume 10, Number 19, pp 39983-39998, 2015
  3. Cheng Zhong Xu, Scalable and Secure Internet Services and Architecture (Chapman & Hall/Crc Computer & I), Chapman & Hall/CRC, 2005
  4. Xu, “Scalable and Secure Internet Services and Architecture”, Chapman &Hall/CRC, 2005.
  5. Mosedale, W. F., and M Cool, R. “Lessons learned administering Nets ape’s site”. Internet Computing Vol. 1 No. 2 (March-April 1997), 28-35.
  6. Yoshilakawa, C., Chun, B., and Eastham, P. “Using smart clients to build scalable services”, Proceedings of Usenix 1997 (January 1997).
  7. Beck, M., and Moore, T. “The Internet-2 Distributed Storage Infrastructure project: An architecture for Internet content channels”. 3rd Int’l WWW Caching Workshop, Manchester, UK (June 1998). http://www.ahe.ja.net/events/workshop/18/mbeck2.html.
  8. Baentsh, M., Baum, L., and Molter, G. “Enhancing the Web’s Infrastructure: From Caching to Replication”. Internet Computing Vol. 1. No. 2 (March-April 1997), 18-27
  9. Colajanni, M., Yu, P. S., and Cardelini, V. “Dynamic load balancing on geographically distributed heterogenous web servers”. IEEE 18th Int’l Conference on Distributed Computing systems (May 1998), 295-302.
  10. Cardelini, V., Colajanni, M., and Yu, P. S. “Dynamic load balancing on web server systems”. IEEE Internet Computing, vol 3, no 3 (May-June 1999), 28-39.
  11. Kwan, T. T., McGrath, R. E., and Reed, D. A. “NCSA’s World Wide Web server: Design and performance”. IEEE Computer, no. 11 (November 1995), 68-74.
  12. Cisco Systems Inc. “Distributed Director White Paper”. http://www.cisco.com/warp/public /cc/cisco/mkt/scale/distr/tech/d_wp.htm, 1997
  13. D. Sanghi, P. Jalote, P. Agarwal, N. Jain, and S. Bose, “A testbed for performance evaluation of load-balancing strategies for Web server systems,” Software—Practice and Experience, vol. 34, no. 4, pp. 339–353, 2004

1,740 total views, no views today

Download PDF File

About the author: admin