If I have two Trino worker servers with different amounts of memory, does the coordinator account for this and send more memory intense work to the larger server?

If, say, I have a worker with 16GB memory and one with 128GB memory, would the coordinator realize this and send most of the work to the 128GB worker?

Answer: When fault-tolerant execution is enabled, the bin-packing is based on the memory estimation of a task + the available memory on a worker, so yes. But when fault-tolerant execution is disabled, the coordinator does not change scheduling based on the memory availability of workers.