People have certainly tried, but I think that these comparisons don't make any sense. What makes the "waste" of crypto stand out is a qualitative concern:
HFT uses energy as a means to an end (i.e. computation), while proof-of-work mining has an incentive structure that directly rewards energy expenditure. In other words, one is energy- (and hardware-)bound, the other isn't.
As a thought experiment: If fusion energy and self-replicating nanobots were to become viable tomorrow, how would that impact HFTs and proof-of-work mining, respectively?
You left it unsaid but wow, PoW is a horrifying grey-goo type scenario that ends in dyson spheres and intersellar war. I had never explicitly thought that through before.
HFT firms use a tiny fraction of the energy of proof of work crypto schemes. Even if you take into account energy requirements they used to build their high speed networks.
The main reason being is that they are all co-located with the exchange, and the byzantine fault tolerance is completely side stepped by the exchange having a single point of serialization between the HFT participants and the matching engine. And then just ensuring everyone has the same length cables to that single point of serialization.
So the policy of everyone connects through this single point, solves the fairness problem.
And then the fault tolerance is often just solved by having an active standby architecture, so if the primary matching engine goes down, you fail over to the backup.
You would be surprised at the low number of servers actually in the primary path for financial exchanges.
Well for the actual trading they will be operating out of the same datacentres as the exchanges which are not particularly massive and will have relatively small limits on the heat (and therefore power). There will be other computers in other bigger cheaper datacentres for analysis and suchlike. And office buildings of course. But not really comparable to e.g. the energy usage of Argentina. I would guess the total is less than one of the massive internet companies like Google or fb but that might be a bit low.
Not sure, but I understand crypto uses > 0.5% of global electricity production. So you can pull up any chart that accounts for the 99.5% of top electricity users and verify HFT is not one of them if that's what you're seeking to show.
lxgr|4 years ago
HFT uses energy as a means to an end (i.e. computation), while proof-of-work mining has an incentive structure that directly rewards energy expenditure. In other words, one is energy- (and hardware-)bound, the other isn't.
As a thought experiment: If fusion energy and self-replicating nanobots were to become viable tomorrow, how would that impact HFTs and proof-of-work mining, respectively?
xyzzy123|4 years ago
posnet|4 years ago
The main reason being is that they are all co-located with the exchange, and the byzantine fault tolerance is completely side stepped by the exchange having a single point of serialization between the HFT participants and the matching engine. And then just ensuring everyone has the same length cables to that single point of serialization.
So the policy of everyone connects through this single point, solves the fairness problem.
And then the fault tolerance is often just solved by having an active standby architecture, so if the primary matching engine goes down, you fail over to the backup.
You would be surprised at the low number of servers actually in the primary path for financial exchanges.
dan-robertson|4 years ago
dataflow|4 years ago