twissandra/twissandra

Twissandra is an example project, created to learn and demonstrate how to useCassandra. Running the project will present a website that has similarfunctionality to Twitter.Most of the magic happens in twissandra/cass.py, so check that out.InstallationInstalling Twissandra is fairly straightforward. Really it just involveschecking out Cassandra and Twissandra, doing a little configuration, andthen starting it up. Here's a roadmap of the steps we're going to take toinstall the project:Check out the Twissandra source codeInstall and configure CassandraCreate a virtual Python environment with Twissandra's dependenciesStart up the webserverCheck out the Twissandra source codegit clone git://github.com/twissandra/twissandra.gitInstall and configure CassandraFollow the instructions for installing and setting up Cassandra.Note, Twissandra requires at-least Cassandra 2.0 or later.And then make sure Cassandra is running:bin/cassandra -fCreate a virtual Python environment with Twissandra's dependenciesFirst, make sure to have virtualenv installed. If it isn't installed already,this should do the trick:sudo easy_install -U virtualenvNow let's create a new virtual environment, and begin using it:virtualenv twisssource twiss/bin/activateWe should install pip, so that we can more easily install Twissandra'sdependencies into our new virtual environment:easy_install -U pipNow let's install all of the dependencies:pip install -U -r twissandra/requirements.txtNow that we've got all of our dependencies installed, we're ready to start upthe server.Create the schemaMake sure you're in the Twissandra checkout, and then run the sync_cassandracommand to create the proper keyspace in Cassandra:python manage.py sync_cassandraStart up the webserverThis is the fun part! We're done setting everything up, we just need to run it:python manage.py runserverNow go to http://127.0.0.1:8000/ and you can play with Twissandra!Schema LayoutIn Cassandra, the way that your data is structured is very closely tied to howhow it will be retrieved. Let's start with the 'users' table. The key isthe username, and the remaining columns are properties on the user:CREATE TABLE users ( username text PRIMARY KEY, password text)The 'friends' and 'followers' tables have a compound primary key. The firstcomponent, the "partition key", controls how the data is spread around thecluster. The second component, the "clustering key", controls how the datais sorted on disk. In this case, the sort order isn't very interesting,but what's important is that all friends and all followers of a user will bestored contiguously on disk, making a query to lookup all friends or followersof a user very efficient.CREATE TABLE friends ( username text, friend text, since timestamp, PRIMARY KEY (username, friend))CREATE TABLE followers ( username text, follower text, since timestamp, PRIMARY KEY (username, follower))Tweets are stored with a UUID for the key.CREATE TABLE tweets ( tweet_id uuid PRIMARY KEY, username text, body text)The 'timeline' and 'userline' tables keep track of what tweets weremade and in what order. To acheive this, we use a TimeUUID for theclustering key, resulting in tweets being stored in chronologicalorder. The "WITH CLUSERING ORDER" option just means that thetweets will be stored in reverse chronological order (newest first),which is slightly more efficient for the queries we'll be performing.CREATE TABLE userline ( username text, time timeuuid, tweet_id uuid, PRIMARY KEY (username, time)) WITH CLUSTERING ORDER BY (time DESC)CREATE TABLE timeline ( username text, time timeuuid, tweet_id uuid, PRIMARY KEY (username, time)) WITH CLUSTERING ORDER BY (time DESC)Fake data generationFor testing purposes, you can populate the database with some fake tweets.python manage.py fake_data <num_users> <max_tweets>num_users is the total number of users to generate and max_tweets is themaximum number of tweets per user. The number of tweets per user is determinedby the Pareto distribution so the number of tweets actually generated will varybetween runs.