Limits hit under native token transfer workload

[mooori]

Setup

• Start a fresh node running [email protected] with config generated by neard init --chain-id localnet.
• Enable memtrie in config.json.

Limits hit

The following limits are hit and limit throughput despite low chunk apply times:

• StorageProofSize limits
• gas/compute limits
• congestion control limits

After effectively disabling these limits by setting them to very large numbers, the following limit is hit:

• produce_chunk_add_transactions_time_limit

Other factors affecting throughput in terms of TPS:

• wait_until specified when sending the transaction. This is probably related to RPC limits.
• state size

Ideas for further analysis

• Repeat with new nearcore releases.
• Analyze the order in which limits are hit.
• Collect data on TPS and chunk apply times before/after removing limits.

[nagisa]

StorageProofSize limits

It would be interesting to see if this continues to be the case for neard built from master with the --features=nightly flag.