[2018-04-21] No, Visa Doesn’t Handle 24,000 TPS and Neither Does Your Pet...

[limek]

No, Visa Doesn’t Handle 24,000 TPS and Neither Does Your Pet Blockchain

When it comes to measuring the speed of new blockchains, the comparison is always with Visa. Despite not being a blockchain, the 24,000 transactions per second Visa reportedly handles have attained mythical status. That figure is unquestioningly trotted out whenever scaling is discussed. In reality, claims of Visa’s throughput, as well as those of emerging blockchains, have been greatly exaggerated.

Visa, Scaling, and the 24k Hoax

Bitcoin was envisaged as a payments system and so it was natural, long before the store of value notion emerged, that comparisons would be made with existing global payment systems. Bitcoin’s early adopters knew that if the technology took off, some time in the future it would need to handle magnitudes more transactions per second than the 7 it could muster. Someone mentioned Visa with their magical 24k per second, and it’s stuck ever since.

Only that figure isn’t entirely accurate. In fact it’s not even remotely accurate. In reality, Visa processes around 1,700 transactions per second, a figure it rarely exceeds. The larger number is the one that Visa claims, and it’s the one that’s usually referenced in comparison to bitcoin and every other blockchain. In theory Visa should be able to handle that volume – in fact it’s been reported that its servers can handle as much as 56k tps – but that’s all theoretical, much like the claimed throughput of new blockchains that can operate at the speed of light in the lab, but significantly worse in the wild. There’s a big difference between operating a testnet on a bunch of Amazon servers and a mainnet distributed around the globe.

Read more: https://news.bitcoin.com/no-visa-doesnt-handle-24000-tps-and-neither-does-your-pet-blockchain/

[1714020414]

1714020414

[1714020414]

1714020414

[1714020414]

1714020414

[1714020414]

1714020414

[bbc.reporter]

Even if Visa is able to handle 24k transactions per second, how much are the costs and who pays for them? Now hold that in mind and try to think about it in a decentralized setting. Who then should pay for the costs of a decentralized cryptocoin network if we want fast and almost free transactions?

[jjacob]

Bitcoin does have a scaling problem as we had seen during 2017. Transaction backlog may have fallen now, but nobody can deny that blocks were indeed running full for prolonged periods of time last year So it is important that we recognize it and start think about scaling solutions. It doesn't matter whether Visa handles 24KTPS or not - Visa is able to handle its real time transaction requirement; something the Bitcoin blockchain was unable to do.

[d5000]

1700 tps could be enough for mass adoption, not only because of Visa's throughput: it would enable ~150 million people to create a (on-chain) transaction per day. Or, alternatively, 4,5 billion people (or households) could do one transaction per month.

If we continue with an approximate transaction size of 220 bytes, 1700 tps would be equal to 224,4 MB blocks (1700 * 60 * 10 * 220). A number which surely will be labeled as "reachable" by the big-blocker fraction.

But also the small-blocker fraction could be happy: If we had 50 sidechains, we would be almost there with today's Bitcoin block size (4 MB Segwit block size * 50 = 200 MB), and the remaining transactions can be handled without problems by Lightning, even if it wasn't the revolution many advocates claim it to be. Think about one sidechain per country - that would be more than enough. (If LN is as successful as some think then a few sidechains would be enough).

So the solution to the scaling problem perhaps isn't so far away.

[1Referee]

Even if Visa is able to handle 24k transactions per second, how much are the costs and who pays for them?

Most of the transactions are related to businesses, so it's safe to say that every Visa transaction deducts a certain percentage of the sale value. It also depends on the service you are using that's based on Visa, which in that case might cost you more, but on average I would say they take around 2% of each transaction. In some cases people purchasing stuff through credit cards have to pay a fee as well, so overall, Visa can earn up to 3-4% of the value in total from one single transaction. And aside from that, there are more types of transactions they heavily earn from, which in some cases can bump the fees to even 5-6%.

Now hold that in mind and try to think about it in a decentralized setting. Who then should pay for the costs of a decentralized cryptocoin network if we want fast and almost free transactions?

Users, obviously. Whether people deal with on-chain transactions or off-chain transactions, block space will always be required, and those who want to obtain a spot in a block will have to pay for it, which again, are the users.

[Carlton Banks]

If we continue with an approximate transaction size of 220 bytes, 1700 tps would be equal to 224,4 MB blocks (1700 * 60 * 10 * 220). A number which surely will be labeled as "reachable" by the big-blocker fraction.

But also the small-blocker fraction could be happy: If we had 50 sidechains, we would be almost there with today's Bitcoin block size (4 MB Segwit block size * 50 = 200 MB), and the remaining transactions can be handled without problems by Lightning, even if it wasn't the revolution many advocates claim it to be. Think about one sidechain per country - that would be more than enough. (If LN is as successful as some think then a few sidechains would be enough).

Or we could just do all of that with Lightning, which scales up orders of magnitude better than sidechains can

Why are sidechains always the answer to transaction scaling for you d5000, are you involved in the drivechain project by any chance? Because the drivechain developers are the only other people that never miss a chance to say "hey guys! Ever considered using sidechains for scaling? You'd only need x sidechains for trillions of users..." etc etc, which is remarkably similar rhetorically to what you keep repeating, every chance you get.

Note: no-one else, except drivechain developers (strangely enough), thinks that using sidechains to scale transactions is a good idea, all other Bitcoin developers think the opposite

[Kakmakr]

The TPS will not matter when Bitcoin takes 50% and more of their market share. If we manage to take that amount of users and the Lightning Network is successful, then their servers will be overkill. Why do you invest in more powerful hardware and software, if some of your users can distribute that work and cost for you. 

They will keep on dumping funds into stronger and faster servers and Bitcoin already has the strongest decentralized network in the world.

[mindrust]

Meanwhile lightning.network:

Scalability. Capable of millions to billions of transactions per second across the network.

Blocksize isn't even worth a debate while the amount of transactions lightning network can handle is basically umm... limitless?

Lightning kept the blockchain as decentralized as possible while removing the limit on the tps. Ain't that the most awesome shit?

[Carlton Banks]

Blocksize isn't even worth a debate while the amount of transactions lightning network can handle is basically umm... limitless?

Lightning kept the blockchain as decentralized as possible while removing the limit on the tps. Ain't that the most awesome shit?

yep.

Latest work from Lightning devs suggests there are channel formation algorithms that can keep the Lightning node network decentralised. The hubs objection doesn't matter any more, and neither do block limits (of course ).

[FrueGreads]

I didn't really understood what was the main purpose of the article. Ok so Visa in theory is capable of 24k TPS, or maybe even those 56k TPS, but in practice it usually only operates at 1.7k TPS. Does this change anything in terms of adoption? Visa operates worldwide, and as a user I experience no problems when using my card, so I guess that the current state of the network is able to handle the current levels of adoption, and it can even scale further. So knowing the exact numbers of TPS doesn't seem that important to me.

Anyway, I do believe that Bitcoin will have no problem handling mass adoption with lightning network, and it will be able to do it in a cheaper and decentralized way. What else do people want?

Actually I do believe that Visa does for banks, is what LN will do for bitcoin. I guess that credit cards are in fact a "side chain" solution, just like LN. If we think about it, when we make a payment using a credit card, the money doesn't move straight out of our account right away. In fact it takes several days for the real transaction to be complete. LN will make the same thing possible for bitcoin, and again, it will be able to do it without the need to a middle man, resulting in even smaller fees, and that will actually beat Visa.

[d5000]

Or we could just do all of that with Lightning, which scales up orders of magnitude better than sidechains can

Why handle everything with Lightning? Lightning has also disadvantages, which are accentuated with higher volume (possible hub exit scam with massive channel-closing, opening/closing bottlenecks etc.). It's also better to have a system with different layers with different characteristics, not only one additional layer, so if something goes wrong with one of them we haven't to fallback to on-chain tx alone.

Why are sidechains always the answer to transaction scaling for you d5000, are you involved in the drivechain project by any chance?

No, although if I had the programming skills, I perhaps would participate - does this already count as "bad intention"?

The attractiveness of sidechains, for me, lie in the fact that the simple and effective blockchain transaction paradigm is preserved, while LN introduces a whole set of new paradigms and some new centralization risks. Additionally, once a sidechain is created and a lot of value was transferred to it, it almost doesn't need to interfere with the main chain anymore, while with LN, opening/closing transactions would be relatively frequent. (The two-way-peg in sidechains wouldn't be used very frequently - but it must work as a possibility so the peg holds.)

I think both solutions can perfectly convive harmonically, so why initiate a battle between them?

[Carlton Banks]

Lightning has also disadvantages, which are accentuated with higher volume (possible hub exit scam with massive channel-closing, opening/closing bottlenecks etc.).

This doesn't exist, there is no "exit-scam" attack. If a hub closed channels, everyone gets their money by definition, that's what channel closing means

It's also better to have a system with different layers with different characteristics, not only one additional layer, so if something goes wrong with one of them we haven't to fallback to on-chain tx alone.

Sidechains aren't a different layer, as they don't use mainchain scripts as building blocks

The attractiveness of sidechains, for me, lie in the fact that the simple and effective blockchain transaction paradigm is preserved,

Adding extra transaction chains doesn't preserve the simplicity of the blockchain paradigm at all, it complicates it. Using Lightning, of course, does not complicate the way the blockchain functions, because it actually is a layer on top of the blockchain

while LN introduces a whole set of new paradigms and some new centralization risks.

Right, the Lightning paradigm is a new paradigm (there's only 1 though, you accidentally made it sound more complicated than it really is). And as for centralisation risks, I think the Lightning devs have got that taken care of with the cyclic routing concept.

Additionally, once a sidechain is created and a lot of value was transferred to it, it almost doesn't need to interfere with the main chain anymore

Lightning just uses time locked transactions to expand capacity. There is no "interfering" with the mainchain, everything Lightning does can be done legitimately on-chain.

while with LN, opening/closing transactions would be relatively frequent.

This is... a mistruth that Lightning critics often use? Are you one such critic, d5000? There's no reason to close Lightning channels once most people are using it, why pay 500 satoshi fee when you can pay a 0.0001 satoshi fee?

(The two-way-peg in sidechains wouldn't be used very frequently - but it must work as a possibility so the peg holds.)

What if the sidechain pulls an "exit scam", lol


Sorry d5000, sidechains can work for non-BTC assets secured by Bitcoin miners. But no-one else in the cryptocurrency development field thinks they provide anything meaningful to help transaction scaling, oh, except the Drivechain devs, who go on and on about it every chance they get

[d5000]

Lightning has also disadvantages, which are accentuated with higher volume (possible hub exit scam with massive channel-closing, opening/closing bottlenecks etc.).

This doesn't exist, there is no "exit-scam" attack. If a hub closed channels, everyone gets their money by definition, that's what channel closing means

I think you know that isn't what I meant.  

The scenario I worry about is: a big hub that has thousands of channels open and "reverses" their state, already knowing that there is not enough on-chain capacity to close all these channels in time. With sidechains as an additional layer this kind of attack would be unfeasible because of the higher capacity.

It's also better to have a system with different layers with different characteristics, not only one additional layer, so if something goes wrong with one of them we haven't to fallback to on-chain tx alone.

Sidechains aren't a different layer, as they don't use mainchain scripts as building blocks

But once established they're independent from the main chain in the sense that they work without needing it. In the worst case, if the peg was broken, they would behave like an alt-coin.

The attractiveness of sidechains, for me, lie in the fact that the simple and effective blockchain transaction paradigm is preserved,

Adding extra transaction chains doesn't preserve the simplicity of the blockchain paradigm at all, it complicates it.

Where is the "complication"? A "sidechain world" can even be hidden from the user, as it is the case with Lightning.

And as for centralisation risks, I think the Lightning devs have got that taken care of with the cyclic routing concept.

Should I really repeat all the arguments of these anti-Lightning critics? I'm not anti-Lightning, But in a blockchain with high on-chain transaction fees - very likely if we only rely on LN and on-chain tx - cyclic routing would not be very usable and very expensive for people without technical knowledge that want it simply "to work". That's why most users - in this high-fee ecosystem ! - simply would use a channel to one hub, very likely their preferred exchange. That doesn't look very decentralized ...

With sidechains as an alternative, fees would not rise that high, and that's also better for Lightning, because cyclic routing becomes more attractive, and the whole system would be more decentralized. The more the Bitcoin ecosystem depends on LN alone, the higher the centralization risks.

Additionally, once a sidechain is created and a lot of value was transferred to it, it almost doesn't need to interfere with the main chain anymore

Lightning just uses time locked transactions to expand capacity. There is no "interfering" with the mainchain, everything Lightning does can be done legitimately on-chain.

That isn't what I wrote.

while with LN, opening/closing transactions would be relatively frequent.

There's no reason to close Lightning channels once most people are using it, why pay 500 satoshi fee when you can pay a 0.0001 satoshi fee?

Why are people then talking about the need of "rebalancing" services (e.g. here)? I'm not saying that you would have open and close channels every day or even every month, but most likely you will need to do it several times a year if you use it frequently, because the payment flows between actors change, and sometimes - above all, if you participate actively in routing - you will be in a situation where flows to one side of the channel are more frequent than the other, which needs to be countered by an on-chain transaction. On a sidechain, you'll very likely never have to interact again with the main-chain.

(The two-way-peg in sidechains wouldn't be used very frequently - but it must work as a possibility so the peg holds.)

What if the sidechain pulls an "exit scam", lol

Who is "the sidechain"? It's not a single actor (if you don't mean a "federated" sidechain like Blockstream's Elements project). All what's possible is an 51% attack. Major sidechains will be as secure as currently major altcoins are.

Why so much hate against sidechains, Carlton? My stance is clear: a Bitcoin ecosystem with sidechains as well as LN would be better, more scalable, more decentralized and more secure than a simple onchain/LN combination.

[Carlton Banks]

1. Cyclic routing is automated, I don't know where you're getting the idea that it needs user intervention or any kind of extra skills
2. That's (probably) the end of the whole "scary hubs" situation, be happy
3. Rebalancing will be handled without services using the "channel factories" concept

You're not up to date on Lightning development, and woefully blinded to the pointlessness of using sidechains as a way to add transaction capacity, it scales the exact same way that adding block space does, i.e. it doesn't scale at all, it's not a scaling concept, it just adds capacity at the same scale.

You also fail to address the major problem with transaction sidechains: if you need to download & validate an entire chain you've not used up until now, that could take a long time if it's several hundred GB. "Just atomic swap across chains" doesn't help you if your coins are on a congested chain, if that was the reason you wanted to hop chains. And what if popular sidechains become mining monopolies? There are all sorts of strategic ways that miners could use multiple chains to screw with users, spamming chain A to drive users to chain B, then spamming B to drive them all to some other. This "uncomplicated" solution sounds like a great way to complicate things, for zero gain in transaction scaling.

[d5000]

1. Cyclic routing is automated, I don't know where you're getting the idea that it needs user intervention or any kind of extra skills

You keep misunderstanding - I was talking about high fees. The lower the on-chain fees are, the less the cost of opening a channel, the better the inter-connection, and the lesser the centralization risk. You need some skills to chose low-fee periods to open channels.

If you want an "one-main-chain-and-lightning" concept, you'll need big blocks to achieve fees that are low enough once mass adoption really has begun. As far as I know you, I think you won't be that happy about that concept. Me neither.

2. That's (probably) the end of the whole "scary hubs" situation, be happy

I don't think so. I'm testing the Eclair beta and there are already some pretty big hubs on testnet. Why do you think that that wouldn't happen if we had mass adoption? Hubs will always be the cheapest option for people that want LN just to work, so if fees are high, then users will probably use a client that doesn't fully-automatic routing but gives them the possibility to connect to a hub. If fees stay low, then we'll have a different situation and the decentralized LN becomes feasible.

3. Rebalancing will be handled without services using the "channel factories" concept

OK, I didn't know about channel factories, will investigate further.

it scales the exact same way that adding block space does, i.e. it doesn't scale at all, it's not a scaling concept, it just adds capacity at the same scale

No, because the sidechain can be maintained by another set of full nodes, so the RAM/CPU resources are distributed. As I said above, ~50 sidechains + LN will probably be all we need for global BTC adoption by humans, so we'll never have scalability problems again. And sidechains do not necessarily have to be merged-mined, so your miner centralization fears may be a (probably minor) problem of one sidechain concept (Drivechain and similar concepts) but not of all of them. Miner centralization on one main chain keeps being the major security risk.

[bbc.reporter]

Even if Visa is able to handle 24k transactions per second, how much are the costs and who pays for them?

Most of the transactions are related to businesses, so it's safe to say that every Visa transaction deducts a certain percentage of the sale value. It also depends on the service you are using that's based on Visa, which in that case might cost you more, but on average I would say they take around 2% of each transaction. In some cases people purchasing stuff through credit cards have to pay a fee as well, so overall, Visa can earn up to 3-4% of the value in total from one single transaction. And aside from that, there are more types of transactions they heavily earn from, which in some cases can bump the fees to even 5-6%.

Now hold that in mind and try to think about it in a decentralized setting. Who then should pay for the costs of a decentralized cryptocoin network if we want fast and almost free transactions?

Users, obviously. Whether people deal with on-chain transactions or off-chain transactions, block space will always be required, and those who want to obtain a spot in a block will have to pay for it, which again, are the users.

Thank you for that reply hehehe. But there is one bitcoin fork that implies that unlimited block sizes are possible to scale the network and keep fees very low at almost zero. Who will incur all the costs in that scenario?

[Alex077]

No, Visa Doesn’t Handle 24,000 TPS and Neither Does Your Pet Blockchain

When it comes to measuring the speed of new blockchains, the comparison is always with Visa. Despite not being a blockchain, the 24,000 transactions per second Visa reportedly handles have attained mythical status. That figure is unquestioningly trotted out whenever scaling is discussed. In reality, claims of Visa’s throughput, as well as those of emerging blockchains, have been greatly exaggerated.

Yes, this is a good clarification, we really believe in some facts and numbers that are far from reality. Of course, Bitcoin is far away to 1,700 transactions per second, but there are other cryptocurrencies that may perform fast transactions in large quantities, and bitcoin can live as a digital gold for store of value.

[franky1]

lets add my 3bits to this

1. visa does NOT store bank balance information on its servers and does not fully process a payment in seconds. here in the UK a bank branch has a sort code(americans have route numbers). what visa does is asks the bank branch server for a balance to ensure customer can fund the payment. and then tells the bank to put the funds on hold. it is then the bank that actually does the payment processing/fund movements

visa's operation is basically the unconfirms tx relay. letting a peer check balance and check the authorisation(signature) is valid..
so with fibre being available i can easily push through 24,000tx's to another node

2. even checking out visa info
"With Visa, you can scale your business to conduct speedy transactions in person or online, knowing the world's largest retail electronic payments processing network is behind you. VisaNet handles an average of 150 million transactions every day"

150m/24= 6,250,000per hour
6.25m/60=104166.6666666667 per minute
1736.111111111 per second

yep averages 1,736 TPS

3. also adding to point one about visa not making a settled payment....., the visa network is not one network
each country has their own separate entity (like a side chain) this is how america pretty much said for visa to stopworking within russia. because visa has a separate network for russia.
so adding it up may look big as in nearly 900million card holders world wide doing 150mill tx's a day..
but thats like saying bitcoin core bitcoin cash bitcoin gold can handle a dozen megabytes of data every ~10minutes by saying they are "all bitcoin"

[Carlton Banks]

1. Cyclic routing is automated, I don't know where you're getting the idea that it needs user intervention or any kind of extra skills

You keep misunderstanding - I was talking about high fees. The lower the on-chain fees are, the less the cost of opening a channel, the better the inter-connection, and the lesser the centralization risk. You need some skills to chose low-fee periods to open channels.

If you want an "one-main-chain-and-lightning" concept, you'll need big blocks to achieve fees that are low enough once mass adoption really has begun. As far as I know you, I think you won't be that happy about that concept. Me neither.

2. That's (probably) the end of the whole "scary hubs" situation, be happy

I don't think so. I'm testing the Eclair beta and there are already some pretty big hubs on testnet. Why do you think that that wouldn't happen if we had mass adoption? Hubs will always be the cheapest option for people that want LN just to work, so if fees are high, then users will probably use a client that doesn't fully-automatic routing but gives them the possibility to connect to a hub. If fees stay low, then we'll have a different situation and the decentralized LN becomes feasible.

Check out the cyclic routing concept too, you're wasting your time writing about what will likely be remembered only as the original conception of Lightning, it will become quickly outdated

[d5000]

Check out the cyclic routing concept too, you're wasting your time writing about what will likely be remembered only as the original conception of Lightning, it will become quickly outdated

We'll see, we'll see. (Didn't I write that I know about cyclic routing?)

In the end, if LN alone achieves a better performance, better security, more decentralization and better prices (less transaction fees) than a sidechain-LN combination, I'll be happy, too. Channel factories look interesting, but seem to be in a very early stage. For now, as all those nice technologies are still being tested, I think it's not the worst option to follow, evaluate and/or support several approaches to the scaling problem. If you think it is - nobody forces you to care about sidechains, you can safely ignore them (and me and my "wasting of time" )