Sign up to receive blog updates in your inbox.
I am excited to join Rockset this week. It was an easy decision because of the people at Rockset, the Rockset product, the pivot for me from OLTP to real-time analytics and of course the new performance problems I get to think about.
I worked with Rockset’s founding team at Facebook on infrastructure projects. They were highly productive and I want to work with them again. We did social graph OLTP at a ridiculous scale via MySQL, TAO and several other projects. As a bonus, I learned that Nathan Bronson, another highly productive peer from my Facebook years, recently joined Rockset.
I am excited by the product - real-time analytics with excellent performance, usability and manageability. The opportunity to work on a cloud-native solution is compelling as it provides new ways to make things better, especially when using an LSM. Two examples are RocksDB Cloud and disaggregating LSM compaction. RocksDB Cloud adapts RocksDB to use object storage like S3. The high durability of cloud object storage lets me focus on solving higher-value problems and worry less about losing data. Disaggregating compaction makes use of elastic compute in the cloud and prevents the variable compute needed by compaction from interfering with things that matter to users — fast queries. All of this leads to a better user experience. Beyond the two examples, the cloud-native opportunity for data-intensive applications is even larger as explained in this post.
This is a pivot for me and I welcome it. My focus since 2005 has been transaction processing via a product, MySQL, that was designed long before cloud-native was a thing. While I will always be a fan of MySQL I am happy to shift to real-time analytics and a cloud-native architecture. Change is good and I look forward to the journey as I learn modern C++, Kubernetes, cloud infrastructure, SQL for analytics and more.
Of course I am excited to continue working on performance, but it is never just about performance. The goal is great performance while also meeting the goals for stability, efficiency, usability and manageability. I look forward to explaining how the RUM Conjecture is relevant for real-time analytics.
I will continue blogging about performance at Rockset, but first I have much to learn so I can be a productive member of the team. I am excited for my Rockset journey to begin!
Sign up for free