June 20th, 2014






“We are happy we decided on Cassandra, it is scaling well, and we are nicely using it in our private cloud.”

– Shamim Bhuiyan, Architect, AT-Consulting

Shamim Bhuiyan: Architect at AT-Consulting




The company AT-Consulting provides consulting services for telecommunications, health and banking sectors. As a architect, my role is to design, research and develop that can drive successful results.


Mobile Network Portability

At AT-Consulting we’ve worked with the largest Russian telephone operator to provide users to change their telephone operator while keeping their own number; commonly known as mobile number portability, or MNP.


Oracle vs. Cassandra

Initially we were evaluating Oracle RDBMS, which is not very good for horizontal scaling and where licensing is costly for such non-transactional data.

Project MNP now implements a lot of business processes: storing non transactional technical logs in Cassandra for future analysis, such as SLA, and business process states, all in real time. We ended up choosing Cassandra based on the following criteria:

  • Replication through various Datacenters
  • Horizontal scaling and no single point of failure
  • Hadoop, Spark integration
  • Very big community


Cassandra and Spark

We are now using Cassandra for storing the non transactional data such as application logs and using Spark and Shark for analysis. We are happy we decided on Cassandra, it is scaling well, and we are nicely using in our private cloud.

In this moment we are using Cassandra in 2 projects. Their deployments look like:

1. 8 datacenters, 32 virtual machines, 33 TB/year and 11,000 transactions/second.

2. 2 datacenters, 8 machines, 6 TB/year and 4,000 transactions/second.

Cassandra is cool and it’s even more cool when you know it’s architecture and data model.