In this PR I prepared the first iteration of katana benchmark. I creates 2000 prefunded accounts, and call move with each of them. Each transaction is run as a separate tokio task to not create a bottleneck on the client side.
Due to the number of needed prefunded accounts there was a need for a slight change in katana, to accept u16 instead of u8 as a number of prefunded accounts.
The best result I got was 917 transactions in a single, 3s block!
Future impovements:
Using katana runner instead of called instance of katana.
Caching generation of prefuned accounts, as it takes about 10s.
Lauching new instance every test to eliminate the need for fetching nonces
Using logs to read number of transactions in each block
matzayonc
commented
10 months ago
In this PR I prepared the first iteration of katana benchmark. I creates 2000 prefunded accounts, and call move with each of them. Each transaction is run as a separate tokio task to not create a bottleneck on the client side.
Due to the number of needed prefunded accounts there was a need for a slight change in katana, to accept
u16instead ofu8as a number of prefunded accounts.The best result I got was 917 transactions in a single, 3s block!
Future impovements: