Why I Joined Clustrix

 

I recently joined Clustrix as vice president of engineering after more than 25 years in databases and information management. Throughout my career I have focused on innovations in parallel database technology and analytics—primarily in building business-critical products for the enterprise—and ClustrixDB is no exception. Here’s why…

I Believe in MPP

I’m a firm believer in MPP (Massively Parallel Processing), or loosely coupled and shared-nothing architecture. Shared-nothing database machines have the best performance. Their dataflow approach to SQL query execution gives both pipelined and partitioned parallelism. Modular MPP architectures enable systems to grow incrementally, so CPUs, memory, and disks can be added either to speed up the processing of a given query or to scale up the system to process a larger query in the same time. The Tandem (now HP) NonStop SQL system demonstrated that it scales almost linearly. It is cheaper to put together an MPP cluster with commodity hardware.

Seven MPP relational databases are available today: Tandem (now HP) NonStop SQL, Teradata, Informix, Vertica (now HP), ParAccel (now Actian), Greenplum (now EMC), Hadapt, and Clustrix. NonStop SQL (the first MPP OLTP database) and now ClustrixDB are the only MPP OLTP databases. The rest are designed for data warehouse purposes. Oracle, IBM (DB2), Microsoft SQL Server, and Sybase are OLTP players, but none of them are MPP. Linear scalability is an issue for these solutions and they rely on ever-more expensive hardware to create scale. Clustrix is the only MPP OLTP database that can handle mixed workloads simultaneously.

The Clustrix Real-Time Analytics Capability is Unique

Unlike the other MPP databases, ClustrixDB can process 3 million transactions per minute and uses consistent hash partitioning of data. This unique capability paves the way for near-liner scalability without impacting production downtimes. ClustrixDB also supports OLTP and OLAP in the same cluster and has done some interesting things in the area of mixed workloads. Tandem NonStop SQL achieved mixed workloads via process priority inversion, while Clustrix has taken a new approach to achieve the same goal. With my experience in OLTP and data warehouses, I am excited to make new contributions to mixed workloads (OLTP and OLAP) in the same cluster and to fuel interest in real-time analytics.

I Believe the Enterprise Wants SQL

The Big Data, Hadoop, and NoSQL waves have been with us for a few years. Users, venture capitalists, and startups flocked to these ideas. However, enterprise users and vendors have now realized that NoSQL and Hadoop are not yet ready for the enterprise in the areas of real-time analytics, cluster management, and configuration. And all those solutions require significant new skills and tools—major inhibitors for enterprises that are not ready to abandon SQL. I believe Clustrix has a proven scalable and mission-critical SQL foundation that is ready for the enterprise, complementing NoSQL and Hadoop to provide real-time analytics on live operational data.

 

– Rao Yendluri