Greybeam and the Layers of Optimization

Arsham and I attended an optimization presentation at Snowflake's San Mateo office last week. Maneesh (VP Value Engineering @ Snowflake) painted the picture of optimization with an interesting analogy that I thought would make the perfect introductory Greybeam blog post.

Consider a jar filled with sand, pebbles, and big rocks, each individual object is an inefficiency within your Snowflake account. Our optimization goal is to empty the jar by removing the contents within (no, we can’t just tip the jar over!). Removing the big rocks represents the low-hanging fruit. There are so few of them, and the impact is so high that spending the time picking each one out one by one makes a material difference. This is akin to getting your warehouse configurations right: sizing appropriately, auto-suspending within a reasonable window of time, etc.

The next layer consists of the pebbles. There are more of them, but they’re still manageable to address one by one. This includes setting appropriate clusters, cluster keys, caching strategies, etc.

Finally, we have the sand. Millions upon millions of grains of sand is the final step in our journey to empty the jar. Each grain of sand represents countless queries running through your warehouse. There are simply far too many to even consider tackling individually. This is why we’re building Greybeam, the last mile in Snowflake optimization. We automate SQL optimization at scale.

At Greybeam, we believe that no one should ever have to suffer through the insufferable pains of Snowflake optimization. We have personally spent countless hours optimizing SQL queries, and our time is far too valuable to be thrown away on mindless optimization. That's why we're on a mission to help our customers reclaim their time, money, and resources, starting with the humble SQL query.