[Cryptography] How fast can a blockchain go ?

[Phillip Hallam-Baker]

On Fri, Mar 2, 2018 at 2:03 PM, John Levine <johnl at iecc.com> wrote:

> Sort of apropos of recent discussions, what sort of performance are
> people getting out of blockchains, both transaction rates and latency?
> I realize there are designs with intermediaries that aggregate
> transactions but if you have to depend on the intermediary you might
> as well depend on someone running something fast like DB2.
>
> Bitcoin is presumably the worst case, 10 tps and latency of 20 minutes
> to an hour or more if you want to be sure your transactions are
> published.


​The real question ​is how fast can you finalize a transaction so that it
is fixed. Blockchain is a silly way to do that because there is a ten
minute gap between blocks being added and the chain does not become
dependable until several blocks are added.

​I have not benchmarked my stuff but the only real limit is how fast the
database can serialize transactions.​ I see no problem getting 10K
transactions a second per node. The number of nodes is essentially
unbounded and a node can cross-link with another at any time.

Hashgraph and others are developing similar structures.


If someone needed a blockchain fast enough to keep up with the NASDAQ, it
would be reasonably straightforward to do. Just spend enough cash. The
transactions only settle at the end of the day and there are intermediaries
who hold liability during the day. So the system is kept in check.

The under appreciated part of BitCoin is the smart contracts. They don't do
much in Bitcoin world because it isn't really connected to anything real.
But if you were doing a system in which party A exchanges money/stock with
B and each then want to trade with C and D, you could use conditional
contracts, this one is good if node X has fixed point Y.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.metzdowd.com/pipermail/cryptography/attachments/20180307/f5fd6c5c/attachment.html>