EOS throughput is only 34 TPS? CTO fires back: That's still twice as fast as Ethereum.
A recent research report from the London Blockchain team pointed out that the current throughput of EOS.IO is only 34 TPS, far below the 3996 TPS as shown on the chain data. In response to this, Daniel Larimer, the CTO of EOS.IO developer Block.one, wrote a special article presenting multiple arguments and questioning the author's deliberate avoidance of Ethereum's flaws.
On June 3, a London-based blockchain team released a paper titled "Revisiting Transactional Statistics of High-scalability Blockchain", claiming that the current throughput of EOS.IO is only 34 TPS, Tezos (XTZ) is 0.43 TPS, and Ripple (XRP) is 15 TPS.
Table of Contents
Daniel Larimer: Author Misuses Throughput Definition
Larimer expressed in a blog post on the 11th that the paper collected a lot of data, but readers must question the completeness of the paper, as it attempts to redefine TPS as "the amount of valuable work a system can process." Larimer believes TPS has three potential definitions:
- How many transactions a system can handle
- How many transactions a system actually processes
- Only counting transactions you deem valuable
Larimer's dissatisfaction lies in the fact that the author clearly calculated the TPS of each project based on the third definition, arguing that evaluating system scalability should not differentiate between "valuable" and "non-valuable" transactions.
Valueless Transactions on the EOS Chain
In November last year, the EOS-based project EIDOS distributed a large number of airdrops, causing 80% of transactions on the EOS mainnet to be related to this airdrop, leading to a surge in on-chain CPU resources, making normal transactions and DApp usage difficult. With EOS's feeless transaction model, the chain is more susceptible to being flooded with low-value transactions like gambling, airdrops, and spam.
Larimer explained:
Scalability, literally, refers to the capability to scale, so we should not concern ourselves with what that capability actually does or how it is used. From this, we can conclude that EOSIO has scalability, and that capability is unrelated to how users use it.
Larimer believes this paper is biased, pointing out that the author had received donations from the Ethereum Foundation in the past while researching Ethereum's transaction fee model, possibly intentionally excluding Ethereum from the comparison in the report. Larimer concluded by criticizing:
In any case, even if the author's definition of TPS is unclear, even if Ethereum is completely unaffected by spam, gambling games, and airdrops, 34 TPS would still be twice as much as Ethereum's.