Sup,<p>I have been collecting tweets for 4 days now, using a app that I haven coding for the last 5 months.<p>The reason why I did this app was because I wanted to make user-based recommendations, and other types of data mining using Mahout, and I didnt find enough data for my experiments.<p>About the app
I am using a single 8GB-ram Centos Server hosted on the Rackspace cloud with a cost of less than 15 dollars per day. It can process up to 100 (90 - 105) twitter profiles per second. It works with a average of 2GB of ram and 90% CPU. Its completely fault tolerant. It can process other social networks as well using a simple parse-template.<p>I was able to collect 90+ million tweets from more than 6 million -- the db has 20MM users -- users using JAVA, memcache, mysql, php (visualization), a non ACID architecture, using a object-like structure (no-sql?).<p>I hope this datasets helps you get into the big data world.<p>The current sql dump is too big (66GB) to put in one of my servers so please skypeme:calufaxp or email me calufa{a}gmail.com if you want the data. BTW, the data is FREE...<p>If anyone has a server where I can upload this sql and let others download it let me know.
Thanks to Jason for putting this up on the archive.org site:
<a href="http://www.archive.org/details/2011-05-calufa-twitter-sql" rel="nofollow">http://www.archive.org/details/2011-05-calufa-twitter-sql</a>
<a href="http://codebiatch.com/" rel="nofollow">http://codebiatch.com/</a> .. the file is still uploading if it's not in there yet. Good luck with your project!
So, you are scraping Twitter (likely violating their ToS) to get users' Tweets (likely violating their copyright) and now posting about it on HN? When Twitter is selling chunks of its stream, e.g. via InfoChimps?<p>I don't want to be mean, but this doesn't strike me as a very good idea.