Michael Novogratz recently made headlines with a bold assertion during an interview with TheStreet that the EOS blockchain is, “already doing 5,000 transactions a second. It should be doing 50,000 transactions per second in a few months.”
Novogratz is not exactly an impartial party. He is the CEO of Galaxy Investment Partners, the venture capital firm that announced a $325 million partnership with EOS creator Block.one at the beginning of this year. But even considering that it’s in his interest to promote EOS, these numbers seem inflated.
The greatest challenge to blockchain adoption is scalability. EOS can’t become a widely used settlement platform unless it’s capable of managing hundreds of times as many transactions as either the bitcoin (three or four transactions per second) or Ethereum (about 20 TPS) blockchains have so far been able to.
Frequently cited as the gold standard of transactions per second, Visa is reputed to handle roughly an astounding 1,700 TPS during normal periods of traffic. But the estimates for Visa’s maximum capacity have been as high as 56,000 TPS. Though there is some debate about those numbers, most people in the space agree that a viable settlement platform would have to process thousands of transactions per second.
Clearly, if Novogratz is correct, it would be huge news. Not only would EOS have vaulted into Visa territory, it could actually be performing three times as many transactions as Visa’s typical transaction rates, and soon, it could match Visa’s most optimistic numbers.
The question is: where did he get those numbers? It’s hard to say. There are several sources of data for EOS traffic, but none irrefutably confirm Novogratz’s proclamation.
According to the EOS meta site, the best day ever for EOS was July 14, 2018, with the number of transactions exceeding 3 million.
The previous day had been the first time daily transactions had exceeded one million, totaling a little over 2.2 million. But these two impressive days look like an anomaly on the chart and are bracketed by periods of significantly lower transaction numbers. July 15 had only 35,423 transactions, while there were a mere 325 on July 12. Bouncing back, there were 2.5 million transactions on July 16, then down to less than half-a-million. There are no numbers available yet for last week.
Unless the past week has had numbers that dwarf EOS’ best transaction rates, it has achieved nowhere near 5,000 transactions per second, at least for any sustained period of time. The available numbers instead show that, on its best day, EOS handled an average of only 37 transactions per second.
For the sake of comparison, 5,000 transactions per second sustained for a 24-hour period would produce a total of 432 million transactions. 50,000 – the number Novogratz claims EOS will soon hit – would total more than 4.3 billion.
But these are averages; it’s certainly possible EOS has handled thousands of transactions in a second, but it has yet to maintain such numbers consistently over a 24-hour period.
Another source of EOS statistics, EOS Network Monitor, offers a statistic that is more on point. That site claims that EOS has in fact hit quadruple digit TPS. As of the writing of this article, the site states the all-time high was 3,097 transactions in a single second. Clearly respectable, but still far short of the 5,000 TPS Novogratz described.
A third data source, Blocktivity, claims EOS handled an average of 257,729 per day (this is an average of only about 3 transactions per second). But, the site also claims EOS is using only 0.02 percent of its total capacity. At one hundred percent, EOS would be handling about 15,000 per second. Blocktivity, however, provides no information of how it compiles its data or how it determines capacity.
So, while none of the data available actually confirms Novogratz’s assertion that EOS is hitting 5,000 transactions per second, the good news for EOS is that it probably doesn’t really need to. After all, Visa (the gold standard of TPS) averages fewer than 2,000 per second but does so consistently.
Courtesy of ethnews