1

I am currently doing some measurements of Tezos' performance, running on Carthagenet. For this purpose, I deploy private testnets of variable size with the following sandbox parameters:

{ "preserved_cycles": 2, "blocks_per_cycle": 8, "blocks_per_commitment": 4, "blocks_per_roll_snapshot": 4, "blocks_per_voting_period": 4, "time_between_blocks": ["10", "5"], "endorsers_per_block": 2, "hard_gas_limit_per_operation": "1040000", "hard_gas_limit_per_block": "10400000", "proof_of_work_threshold": "-1", "tokens_per_roll": "8000000000", "michelson_maximum_type_size": 1000, "seed_nonce_revelation_tip": "125000", "origination_size": 0, "block_security_deposit": "512000000", "endorsement_security_deposit": "64000000", "baking_reward_per_endorsement": ["1250000", "187500"], "endorsement_reward": ["1250000", "833333"], "cost_per_byte": "1000", "hard_storage_limit_per_operation": "60000", "test_chain_duration": "1966080", "quorum_min": 2000, "quorum_max": 7000, "min_proposal_quorum": 500, "initial_endorsers": 1, "delay_per_missing_endorsement": "1" }

On every node, I also run a baker, an endorser and an accuser. Moreover, all nodes have the same stake and therefore the same probability to be elected as baker in a certain period.

I was generally happy that Tezos run highly stable and I managed to get a throughput of > 500 tx/s in a 16 and 32 node network with 2 CPUs each. However, for 64 nodes things changed dramatically, and I only managed to get about 3 tx/s. I have already tried to increase the time_between_blocks parameter to ["20", "10"] to account for a larger network size and therefore longer block propagation. I guess that I have to adapt other parameters as well for larger network sizes. Any help would be very much appreciated!

0

Just a remark: as the end of cycle is quite computationally heavy, an idea would be to increase the number of blocks per cycle, to say 256 blocks. It then makes sense to increase blocks_per_commitment and blocks_per_roll_snapshot accordingly.

| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.