Desarrollo de una infraestructura de redundancia para servidores Proxy GNU/LINUX en la intranet de la Facultad de Ciencias.

Abstract

A study of the infrastructure in the School of Science of the Polytechnic School of Chimborazo prior analysis infrastructure redundancy was done. The methodology was quasi-experimental and considerations were made based on the establishment of the availabilit y of internet in the intranet also by studying the stops which were classified as planned and unplanned. Based on th e average service time and mean time to repair the fault the availabilit y was calculated. It was proposed an infrastructure redundancy by implementing two servers, one active and the other as a liability. The team offers the servic e of active and passive internet when the asset does not provide the service, becoming th e passive into active. It was done automatically with soft ware redundancy. As a measurement tool failure time, a bash scriptencoded was used. The results obtained in the investigation determined that the infrastruc ture redundancy can`t prevent failure. The average time of operation of the equipment is the same wi th or without infrastructure and the failure times regardless of fault were reduced. Because permanent infrastructure redundancy checks if the active server is serving and if it fails, the passive server takes responsibility for the service, a time of 1.14 minutes was obtained less than the non redundant infrastructure which its time was 7.5 to 50 minutes. It was concluded th at redundancy is a viable solution implemented and the weak point was identified as the active node in case of failure and the untaken actions. It was also improved the drop time at stops planned in 92.08% and in unplanned in a 89.36%. The average downtime was 90.72%. In terms of availability, it is presented an improvement in a planned downtime in 0.00582% and in the unplanned 0.00576% giving an average overall improvement of 0.00579%. To test the hypothesis, a statistical analysis with t -student method was identified since the number of variables to be analyzed was under 30. It was based on the comparison of means for independent testing. Null hypothesis was determined as the availability of normal infrastructure redundancy infrastructure in which there was no difference. As alternative hypothesis it was proposed an infrastructure availability which is less than the value of redundancy infrastructure. Finally, the null hypothesis was rejected because the significance value is less than 0.05, which was interpreted as greater availability redundant infrastructure for proxy servers GNU / Linux on the intranet of the School of Science.