Before migrating Ethereum Dapps from L1 to L2, it’s important to know the change in gas fee in advance.
Ethereum gas tracker provides real-time gas fee cost.
But Arbitrum, with different mechanism and fee status, use ArbGas along with L1 fixed cost and calldata cost to compensate the chain’s validators for their expenses.
— L1/L2 bridge
Bridging costs much more than L1 transfer, we will analysis later.
— L1/L2 message
For example, the L1 calldata in tx 0x9b3b84 fixed cost(inclusion to Layer 1 inbox) is 0.00049 ETH(1.93$ at 3936$/ETH), calldata(transactions are written on Ethereum) paid is 0.00030 ETH(1.18$ at 3936$/ETH), which are 3.11$, while L2 computation costs 2230 arbgas = 0.0000023 ETH = 0.009$, so 3.119$ in total, almost the same as 3.1314$ in the explorer.
As you can see, the ETH data inbox/calldata cost the majority of the fee and are very expensive.
ps: this may be the reason that ZkPorter uses off-chain data availability. But op’s security relies heavily on transaction emulation based on data availability, much more than zksync, so this could be a disadvantage.
— L2 transaction
The detailed AVM cost is listed here, which use ArbGas=1/100 million* 1 second of CPU time on the Offchain Labs developer laptops in early 2020, with different trade-off(for emulation) compared with Ethereum Gas. So dapps can re-design their code to reduce L2 cost, but as previous mentioned, since the L1 costs more, there is even no need to optimize this L2 code.
One important difference is the speed limit, which requires validators to check one’s assertions, so the VM cannot progress faster than the other’s emulating the execution. This is one disadvantage compared with zk rollup, which has constant and fast verification time, rather than re-execution.