Making Transactional SQL on HBase Scale
If you haven’t tried building a scalable, relational database that can run mixed transactional and analytical workloads, the Splice Machine team has a word of advice: it’s hard.
At HBaseCon 2018, Splice Machine co-founder and CEO, Monte Zweben, took the audience through the company’s journey in scaling out to meet the needs of a large financial services company.
The “battle scars” earned through deploying and scaling the Splice Machine database for this company have resulted in a real-time searchable data store to power groundbreaking levels of customer service. The specs of the project go beyond what other NoSQL or SQL databases are capable of, while scaling out on commodity hardware instead of scaling up on top-of-the-line hardware.
- Contains rolling 2 years of records
- Adds 2 billion records a day
- Searches across 7-8 petabytes and 500+ billion records
- Contains tables with 700+ columns and 7 indexes per table
- Runs 2 million queries a day
- Delivers 98+% of responses in under 2 seconds for complex queries
- Deals with transactions 24x7x365 and is never idle
- Handles ingest and concurrent queries at the same time
- Protect against downtime with a fault-tolerant, hot failover platform
There aren’t many use cases in the world quite like this one, but with the lessons learned, Splice Machine is more capable of solving complex application data platform challenges that you may have.
If you’re looking to offload or migrate workloads from traditional databases and data warehouses, operationalize a Hadoop data lake, or build real-time machine learning applications, sign up for a demo of Splice Machine today from a solution specialist. You’ll see firsthand how having a data platform that is a SQL RDBMS, data warehouse and machine learning platform can simplify data infrastructure, eliminate data movement and reduce latency while powering intelligent applications that can operate, analyze and predict in real-time on real-time data.