Our Big Move Into The Open-Source Database World
July 18, 2016
Developers rule! We know they influence technology decision-making both with their vote in the enterprise and their voice as a community member. When we asked developers what their experience was with Splice Machine, we learned something really meaningful that has changed the way we’re building, marketing and selling our software.
They would hear about us, visit our website, get excited about the mixed OLTP/OLAP workload concept, and with skepticism say, “I’ll believe it when I see it”.
Then they tried the “standalone” edition. It downloaded quickly, installed flawlessly, and ran full SQL including transaction rollbacks. “Wow!” they said. “Maybe this is for real, I want to try this at scale”.
But this is where we lose them.
To get a cluster version, the first person they would interact with was an experienced sales person trained to qualify opportunities. “What’s the use case?” “How will you make a decision?” “What is the budget?” “Who’s your boss?”
“All I wanted to do was see if your cluster version would scale. Let’s talk another time”.
Imagine if Splice Machine made it easy for them to load their data and test the product without any hard questions?
We have learned that the first steps towards building community are to remove friction and to be transparent. As a result, I’m happy to announce Splice Machine is now open source with an incredibly easy-to-use sandbox on Amazon.
So today, that same developer will find:
- Our docs and source code GitHub repository are one click away.
- With just a few clicks, they can build a cluster on Amazon Web Services and have a live Splice Machine instance ready to try in minutes in a sandbox.
- The first call they get from us will be somebody confirming they had a good experience and answer any questions they have. Our qualifying questions will wait.
We are now offering two editions of our new open source product: a Community edition and an Enterprise edition. The Community edition will be free with unrestricted SQL. Developers will be able to test and deploy the Community Edition at scale. The Enterprise edition will exclusively have DevOps features such as backup, encryption, and security, plus future features like Oracle PL/SQL wire-compatibility (Fall) and active-active replication (in 2017). Support comes with the Enterprise edition only.
We are very excited to make the transition to open source and build a larger, more interactive community around Splice Machine. Our team is eager to take suggestions and answer questions from the developers, data scientists and IT staff who want to try our RDBMS and see how it can address their needs. We are excited to share the roadmap with our community, and welcome their ideas and participation. For example, we plan to:
- Improve our physical encoding of data to require less serialization/deserialization, perhaps leveraging Spark’s Project Tungsten
- Improve our transaction engine to make fewer RPC calls to store values and timestamps
- Vacuum MVCC versions more effectively
- Cache Spark RDDs constructed from HBase table scans until “dirtied”
- Build a columnar fractured mirror of Parquet files upon major HBase compactions
We are dedicated to building Splice Machine’s community and will become much more active in related communities like Apache HBase and Apache Spark.
I think you will find the way we combined HBase and Spark to handle mixed OLTP/OLAP workloads very interesting and look forward to hearing how you would make it better — or even better, I hope I get a chance to review your code that actually made it better!