Website Scraping/Spidering - Speed it up!

I am looking to get some listing information about each advert from a site that has tens of thousands of adverts. I have a traditional scraping program that loads each page, reads the data and stores it into a database- which works well. The problem is speed. It takes a long time to scrape the website. The delay is the time to access each page. The program runs through one advert (one page) at a time, taking around 2.6 seconds each. 50,000 adverts then means 36 hours! If I want the photos of each advert as well, then it takes many times longer. I want to speed the whole process up so that it runs in 10-30 minutes or less. My question is how do I accomplish this? Is it even possible? Some things I can do are: 1. run the program on a fast dedicated server 2. get a faster internet connection 3. split the task among a few different servers, each doing a different segment of the site to scrape Doing this will speed things up and I think I could probably bring the scraping time down to 5-6 hours if I use four servers. This would be expensive and I still would not get down to the 30 minute target. What I need is a different approach. Maybe re-write the program to run multiple parallel threads? Maybe there is spidering technology that search engines use that would be more efficient than scraping? Maybe there is something out there that I know nothing about that can address this issue? I am interested in hearing from anyone that knows how to do this, what suggestions you might have and if you have the skills to do it. If you have a solution that allows me to get the data from 50,000 adverts in 30 minutes then we can draw up a project. I will give you the source code of the existing program and all other details and you can then quote me on it. Looking forward to your responses.