Back to search

Democritus University of Thrace (DUTH)

Αποδοτικοί αλγόριθμοι εξισορρόπησης φορτίου στο διαδίκτυο με βάση το περιεχόμενο των αιτήσεων και της κατάστασης των εξυπηρετητών

Abstract

dc:description

High Utilization of web based applications and implementation of more and more services over HTTP, lead to re-design and building of more efficient web servers and services or even implementation of different scheduling policies at requests. The emerging web 2.0 technologies and tools, followed by Google and many social, e-health, e-learning and e-government web services, is a mature indication that the World Wide Web and more specifically HTTP protocol services traffic is definitely becoming the most excessive form of traffic for routers, that shall geometrically increase web servers processing efforts over the following years. Focusing on further improving web server capabilities, distributed web architectures have been developed, consisting of web server farms and load balancing architectures were deployed for HTTP traffic. Such distributed web architectures are the cluster-based web systems. The front end node of a cluster-based web system is the web switch. The web switch receives all in-bound connections that clients send to the VIP address and routes them to the system web server nodes accordingly. A clusterbased web system design is comprised of two basic structural parts: The balancing process, which selects the best suited target servers to respond to requests and the routing mechanism, which redirects clients to the appropriate target server. Cluster-based web systems routing mechanisms are divided into content aware (layer 7) and noncontent aware (layer 2/3/4), while load balancing algorithms used at a web switch are classified into five distinct categories: 1. Stateless non-adaptive, 2. Stateful non-adaptive, 3. Stateless adaptive, 4. Stateful adaptive and 5. Content aware. As stateful/stateless algorithms, we characterise those algorithms that keep track or not of client connection requests. As adaptive/non-adaptive, we characterize those algorithms that take into account web server status metrics feedback and adapt their behaviour based on metric transitions. Content aware algorithms extend adaptive algorithm potential by also investigating HTTP request header information and HTTP request payload size (content length) for balancing decisions. This thesis presents an Adaptive Load BaLancing algorithm (ALBL), for cluster-based web systems. ALBL is and adaptive load balancing algorithm that focuses in prediction of network conditions and web servers’ load without the use of feedback information obtained by web servers. That is achieved with frequent service and network probes towards web servers by the web switch. Tests of performance and scalability of the ALBL implementation against known balancing algorithms used at web-farms are presented and show that ALBL outperforms these algorithms and scales well as the number of balancing servers and web cluster’s contention increases, under specific network or processing load conditions. Based on ALBL capabilities from experimental results, a new content aware load balancing algorithm was also designed; called ALBL/HSC. This algorithm maintains classes of HTTP traffic and differentiates traffic based on content of HTTP requests. Then it utilizes ALBL algorithm to load balance requests per class accordingly. That is, ALBL/HSC maintains separate ALBL processes per HTTP service class for balancing traffic among web servers assigned to each class. ALBL/HSC algorithm was put to the test against ALBL over a cluster based load balancing system. Additionally, CPU performance tests performed at the web switch and performance tests at the web servers indicate both capabilities and drawbacks of content aware ALBL/HSC and adaptive ALBL algorithm. New capabilities that can be part of content aware algorithms are also presented. That is, clustering prediction capability, bandwidth and web farm utilization estimation and sharing mechanisms among classes of HTTP traffic. Finally, in order to improve performance of ALBL algorithm in the ALBL/HSC content aware mechanism with a more adaptive, multi-metric algorithm that receives feedback from both the web switch as well as web servers, a new load balancing algorithm, called AWLB for cluster based web systems was proposed. This algorithm maintains agents both at the web cluster and web servers in order to adapt its balancing decisions accordingly. AWLB algorithm also maintains its own protocol specification for signaling purposes among web switch and cluster nodes and also utilizes other protocols such as SNMP and/or ICMP for its balancing process. Performance gains are shown from tests of AWLB against known balancing algorithms used by web-farms. CPU performance tests at the web switch show both advantages and drawbacks of complicated but accurate adaptive decisions over less accurate but misbalancing policies.

Degree

thesis:*
Grantor dc:publisher
Democritus University of Thrace (DUTH)
Year dc:date
2012

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Κοντογιάννης, Σωτήριος
  • Kontogiannis, Sotirios

Subjects

dc:subject × 20

Rights

Language dc:language
gre

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/28350

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Κοντογιάννης, Σωτήριος; Kontogiannis, Sotirios. Αποδοτικοί αλγόριθμοι εξισορρόπησης φορτίου στο διαδίκτυο με βάση το περιεχόμενο των αιτήσεων και της κατάστασης των εξυπηρετητών. Democritus University of Thrace (DUTH), 2012. http://hdl.handle.net/10442/hedi/28350