In today's testing we had a lot of orders in the system which yielded at some point ~400 settlements. Currently the driver loop simulates all of them which can take a long time even when we do it in batches.
A better solution could be to order by objective value first and only simulate the first N (one batch) settlements. Only if all fail we continue to simulate.
Relevant logs are are from 2021-09-24 14:00 UTC staging solver mainnet. We can also try testing our system with a large amount of orders or rinkeby or xdai.
In today's testing we had a lot of orders in the system which yielded at some point ~400 settlements. Currently the driver loop simulates all of them which can take a long time even when we do it in batches. A better solution could be to order by objective value first and only simulate the first N (one batch) settlements. Only if all fail we continue to simulate.
Relevant logs are are from 2021-09-24 14:00 UTC staging solver mainnet. We can also try testing our system with a large amount of orders or rinkeby or xdai.