@Jackie Can you shed some more light on how StarTree conducted the performance tests that yielded the impressive join query latencies that you presented today? Was TPC-DS or an internal benchmark used? Congrats and thanks!
j
Jackie
09/29/2023, 7:54 AM
We went through all these optimizations during a POC for one use case. The goal is to reach sub-second latency
r
robert zych
09/29/2023, 7:59 PM
Sub-second query latency on a 10TB data set 🤯
j
Jackie
09/30/2023, 12:08 AM
We achieved that by throwing hardware (16 big servers) 😛
r
robert zych
09/30/2023, 12:47 AM
Developing optimizations that reduced the cost of the shuffling also helped significantly.
robert zych
09/30/2023, 12:55 AM
But those optimizations expect the tables to be partitioned in the same way and on the same servers