BrandsEye is an online monitoring and insights tool. We track conversations on social media (Twitter, Facebook, Instagram, Tumblr etc.) and websites and derive business related insights from the data. We help companies improve their bottom line by spending money on the right things, monitoring performance of campaigns, understanding their communities etc.. I am CTO of BrandsEye.
We needed a scalable data store with no single point of failure and excellent write performance. With Cassandra we can easily add more nodes to our cluster as we need more space and/or performance. Being able to tune the consistency requirements for every read and write is a great feature.
We also considered Riak, but Cassandra appeared to be a better fit for our given needs.
We currently store 5-6 million tweets in Cassandra every day as well as millions of processing records from our mention pipeline. We run analytics against this data, using Java and Groovy with a low level CQL code and Apache Spark. Our cluster consists of 3 nodes each, 8 cores and 32GB RAM. With a total storage space of 21TB.
Cassandra is not a relational database even though CQL makes it look like one! Get to grips with the data modeling side, and understand how things are stored in Cassandra.