Hiroyuki Nishi Project Manager at Yahoo! Japan
Yahoo Japan Corporation is a Japanese Internet company providing the largest portal site in Japan, and I’m a manager of a NoSQL infrastructure team.
Search for a modern database
In 2012, our existing products had met issues with linear scalability and operation flexibility. On the other hand, there was a growing need for big-data capable databases throughout our company. We began looking for a modern, open-sourced distributed database as an alternative to our proprietary product.
At that time, my team members evaluated and compared some NoSQL databases such as: HBase, MongoDB, CouchBase and Redis.
As a result of some benchmarking, we found that Cassandra has the lowest request latency, and highest scalability among those databases.
Cassandra at Yahoo! Japan
Today at Yahoo! Japan, Cassandra is used for several services including our eCommerce, finance, maps and our ad delivery systems. Furthermore, many projects are planning to adopt Cassandra in the future.
We have adopted Cassandra 1.1.5, 1.2.15, and 2.0.5 in production. We currently have 3 clusters. One of them is shared by some services through HTTP proxy API, and the rest of them are dedicated to a single user. All of the clusters consist of roughly 30 – 40 nodes.
For now, many of these services use Cassandra to store customer activity information. For example, letting users view their recently viewed item list on the eCommerce portion of our site.
Overall, Cassandra has enabled us to accommodate many in-house demands of large scale data quicker than ever before.
When starting out, take time to learn the data structure of Cassandra, and the differences from a relational database management system. If you understand them, then Cassandra will greatly help you.