printed on December 3, 2009

This can be truly the brief kind of my actual master thesis proposal, that’s attached in PDF format .

Introduction

My bachelor thesis involved making Drupal websites load faster. 80 to 90% within the response time (as observed using the finish user) is used on installing the components from the site. For this reason this really is really the part where optimizations contain the largest effect.

So that you can prove the positive impact of optimizing the loading within the areas of an internet site — therefore showing the task I’d did was an optimistic impact — I researched existing page loading profiling tools. Episodes (meaning various episodes within the page loading sequence) demonstrated up in this region as being a apparent champion.

Also incorporated inside my bachelor thesis, I authored an easy Drupal module 1 that may create simple charts to look for the typical page load time every single day per geographic region.

Despite its apparent (intended) insufficient optimizations, it had been sufficient to exhibit that File Conveyor 2 when integrated obtaining a Drupal site and so offering CDN integration for that site, was an optimistic impact: test site consistently loaded about two occasions as quickly. designed for visitors with slower online connections, for example visitors from Latin america. Without one proof-of-concept implementation, I’d not need had the chance to demonstrate the positive effect on performance.

Context

Increasingly more additional information mill getting to cover focus on page loading performance. Notable recent proposals include SPDY (a suggested re-creation of HTTP.

with far better performance characteristics), Resource Packages (zipping several resource files into one package, to lessen the amount of demands), Web Timing (a suggested specs to integrate parts of Episodes’ functionality to the browser, to enhance better and even more complete measurements).
To complete it, Yahoo is nearly prone to include page loading performance (“page speed”) as being a ranking factor (they’ve already incorporated it running a business owner Tools and they are supplying a faster DNS service, Google Public DNS ).

Problem

Simply applying all known methods isn’t enough, because having a CDN might accelerate your website for half your prospective customers and slow it lower for your better half — although that’s an strangest scenario. That’s that you should manage to do Continuous Profiling (cfr. Continuous Integration ).

Continuous Profiling implies that you’re continuously monitoring your real- world page loading performance: you have to track the page loading characteristics of each loaded page! That alone is easy: precisely what it takes should be to integrate Episodes along with your website. The particular problem is founded on analyzing the collected data. So that you can draw significant conclusions inside the collected data, we have to apply data mining techniques furthermore to visualizing the conclusions which are found.

Precisely what For me is required, is really a factor like Google Analytics. but in addition for page loading performance instead of just page loads .

Proposal

To make certain that is what my proposal is: an analytics suite for tracking page loading performance. A credit card applicatoin that may instantly extract conclusions from Episodes logs and visualize them.

Update — recognized!

Great news, my master thesis proposal remains recognized! My promotor will most likely be professor Jan Van family area Bussche. whom could be a well-known investigator and a very good speaker. He’s many, many publications on query optimization, data mining and related fields in theoretical it.

Following this semester’s exams (which is within the month from the month of the month of january), we’ll feel the details. In case you’re To find out more, understand the full version in PDF format . (The LyX file can also be available.)

File Conveyor may be the daemon that people authored to immediately sync files for the CDN. regardlesss within the file transfer protocol used.&#160&#8617