PepperVillage.com launches with 20 million + nodes

PepperVillage.com (http://www.peppervillage.com) is an Internet comparison shopping, cashback and rewards community website. It is built on Drupal, and starting off at over 20 million nodes and 150 million tags, already represents one of the largest startup sites to run on Drupal.

Through PepperVillage.com, we offer price comparison on millions of products in the USA and the UK, from thousands of online shops through various networks such as Commission Junction and Webgains. The commission payments from these networks for purchases made by registered users of PepperVillage.com are shared back with the users, creating a compelling cashback proposition.

PepperVillage.com is run by Pepper Global Ltd, a London, UK, based startup, and the website has been built by Stragure Software Technologies based in Bangalore, India.

We started work on the site in April 2008, and Drupal was our platform of choice from day 1. At the time, though, some key modules were not available for Drupal 6, and we chose to build on Drupal 5 instead. The rich core functionality coupled with the actively maintained community contributions allowed us to focus on building the bespoke components for this unique proposition. The landscape is very different today, with Drupal 6 offering a richer set of functionality and modules. In the coming weeks, we will be looking at importing the website into Aegir, and running an upgrade to Drupal 6.

PepperVillage.com is available to shoppers in the USA and the UK. The website is also available on mobile using the Switchtheme module with the Browscap patch, running a simple low graphics theme.

PepperVillage.com builds on a number of performance enhancing work that has been done on Drupal.org, and uses a MySQL database cluster configuration, as well as the Apache Solr search engine. The architecture implements APC cache. A number of existing Drupal modules were used straight out of the box, while some bespoke ones had to be built to deliver the commission and cashback calculations and payouts. The affiliate offering is run using Ubercart, which is a significant overhead at this time, but will soon be vindicated once we start maintaining on-site product catalogues and e-commerce.

A couple of significant challenges faced in building a website of this scale are worth specific mention. The first, of course, was the sheer number of nodes. The nodes tables have been flattened out and indexed for better performance. The second challenge was the fact that there are data updates running on the website 100% of the time. Merchant catalogues change all the time, and update operations (Add/Modify/Delete) are happening on the node table and some user tables constantly. The data is updated at the rate of many thousands of nodes per hour, and this needs to be done without causing downtime, even as users might be inserting or updating records in the same tables. PepperVillage.com achieves this through stored procedures, with a master-master replication setup, with separation of automated updates and user initiated writes, as well as read operations. Product feeds are downloaded from partner sites and inserted into a staging database, from which stored procedures create nodes and vocabulary terms. From a hardware perspective, the website is hosted on a cluster of Dual Core quads with 8GB RAM each.

The site receives around 3000 visitors a month at this time as our public launch activity is very recent. This is expected to scale to around 500,000 a month over the next six months, as we implement an Aegir powered white label affiliate solution.As the traffic scales, we intend to keep at the cutting edge by implementing Resin/Terracotta to span servers and ensuring high availability.