Scaling Bitcoin

9 min readSep 15, 2015

There are two competing visions at the core of the block size debate: settlement versus electronic cash.

Settlements

The first vision is that Bitcoin (the network) is a global settlements system, with bitcoin (the currency) as the functional equivalent of gold.

Bitcoin payments on the blockchain serve to settle accounts resulting from retail/commercial transactions, managed by off-chain networks or third parties. These settlement amounts would be high value and may represent the sum of thousands or millions of transactions that take place outside of the blockchain by a broad base of users. This structure closely mirrors today’s financial system, where financial institutions and corporations use exclusive domestic and international networks to settle payments.

Adherents to this vision dismiss the long-term use of blockchain transactions for retail payments (i.e. buying coffee), citing layer 2 networks (such as the Lightning Network or LN) as a a more scalable (and cheaper) environment to conduct this activity.

Electronic Cash

The second vision is that Bitcoin is both a settlement and an electronic cash network. According to this vision, the neutrality of Bitcoin payments on the blockchain are most important. In other words, retail/commercial transactions are made and processed on the blockchain alongside settlement payments.

This vision balks at the prospect of consumers being priced out of the blockchain, as the result of constraining the transaction capacity of Bitcoin, by prohibitive transaction fees that: a) only financial institution or corporations could afford, or b) are comparable to fiat payment processors.

Security and Governance

The debate has largely focused on the security implications and trade-offs of either vision. The sharp division is ostensibly due to the ‘correct’ answer being a subjective valuation of acceptable future outcomes. However, the heart of the debate is the clash of visions I highlighted. This has precipitated a governance crisis.

The purpose of this article is not to argue for or against the merits of one vision over another. Rather, I’d like to focus on Bitcoin’s transaction capacity to service either vision and the implications for the network in the future.

This article follows the example of an excellent post on BitcoinTalk. Below, I compare Bitcoin to popular settlement networks, remittance, and retail/commercial payment processors to discover the practical implications for Bitcoin to compete with any one of these systems.

Bitcoin, Settlement Networks and Payment Processors

To compare transaction capacity, I have selected representatives from retail/commercial payment processors, remittance networks, clearing houses and settlement systems.

Assuming that Bitcoin transactions are on average 520 bytes (incidentally they would be around this size or slightly higher for LN or OpenBazaar multisignature transactions), I calculated the minimum block sizes required to support the transaction volume of these networks:

*Transaction (Tx) capacity for Bitcoin assumes a block size limit of 1 MB. Theoretically 7 tps are possible; practically it is 2 to 3 tx/s. Transaction fees are assumed to be 0.0001 BTC.*

Charted another way:

Notes & Observations

I was surprised to see how close Bitcoin is to competing with the transaction capacity of the Fedwire Funds and TARGET2 systems. Interestingly, both of these systems service a small and exclusive network of financial institutions. However, unlike Bitcoin, both systems are truly inaccessible to consumers.
Automated Clearing House (ACH) is the quintessential clearing house and settlement network for electronic transactions, in the United States, between depository institutions and the central bank. I estimate ~90% of these payments to be non-coffee related. Under the ‘Bitcoin as a settlement network’ vision, the transaction capacity would require block sizes to be > 200 times then what they are now.
ACH only represents the United States!
SWIFT isn’t a settlement network, rather it sends payment order messages that are used to make settlement transactions.
Western Union has ~15% market share in global remittances.
Currently, all of the accepted BIPs do not scale the block size to Visa-level transaction capacity (not even BIP101).
One should factor that Bitcoin will be used for a combination of these applications, until fees make certain use-cases economically irrational (due to protocol changes or soft-limits imposed by miners with a majority of the hashing power). The implications of this means that there will always be transaction pressure to make direct on-chain payments where possible.

Implications

For Bitcoin to fulfill the vision of a global settlements network, with any meaningful market share, the transaction capacity needs to be raised.

For Bitcoin to disrupt international remittances, the transaction capacity needs to be raised.

For Bitcoin to replace the settlement network used by central banks and financial institution, the transaction capacity needs to be raised only marginally, provided fees are sufficiently high to exclude low-value payments on the blockchain.

For Bitcoin to be used as an electronic cash system, the transaction capacity needs to be significantly raised.

Block size calculations again assume an average transaction size of 520 bytes. Cells highlighted in yellow fall under the current block size limit.

The table above shows the transaction capacity and the subsequent block size required to make on-chain settlement payments, at various annual rates, under 3 models:

The estimated number of Bitcoin users by 2019
The rough population of the United States
The world population.

Note that these calculations do not count or attempt to estimate the transaction volume of commercial, financial, or government institutions.

Depending on the efficiency of layer 2 networks (e.g. LN), users may only be required to make (on average) 2 on-chain transactions to open and close payment channels per year (perhaps optimistic). If current projections hold for the number of Bitcoin users by 2019, and the LN is established by that time, then the transaction volume and block size may significantly decrease — even below current levels. If so, miners may experience a non-trivial loss in transaction fees as a result.

However, with greater adoption of Bitcoin, there is a long-term secular pressure to increase the block size. Even for a conservative application of Bitcoin in global settlements, with low transaction volume per person per year, the transaction capacity needs to be raised well above the upper limits of most block size proposals

This means larger blocks are inevitable. Unfortunately, this means that data-center mining and full nodes may be inevitable too depending on hardware and network infrastructure performance over time. The centralization risk depends on the adoption curve, the degree that layer 2 networks remove transactions off-chain, and number of on-chain transactions that are attempted despite the presence of layer 2 systems.

Failing to raise the transaction capacity centralizes the number of actors with access to the blockchain, recreating a financial system that presently exists. Ironically, while the network may be able to run on anyone’s computer, none of the participants may be able to afford to use it. This scenario describes the left side of the Bitcoin failure bathtub.

Lies, Damn Lies, and Linear Trends

When will the block size be a problem?

If we were to look at the linear trend for the past 2 years (from today; the 14th of September 2015):

First off, the R^2 is far from ideal so take this with a large lump of salt:

A rough calculation is that blocks will be full (on average) ~1067 days or ~3 years from today. Note that this is when the purple trend line hits the 1 MB mark. In reality, most blocks will be full well before then.
Also there is no reason to assume that this trend will hold, especially as new projects start to come online (like OpenBazaar) that give users more reasons to transact with Bitcoin. Similarly, an expected increase in the price of Bitcoin could catalyze Bitcoin adoption and transaction volume.
If we calculate the standard deviation of the daily average block size over the 2 year data, it comes up at ~0.12 MB. This means that we may start to notice capacity problems as early as ~810 days from now.
If we also count the 0.75 MB soft limit, along with the block size standard deviation, assuming it isn’t raised by most miners, we’re in for choppy seas in ~310 days. Some mining pools like BitFury are presently mining 900+ KB blocks, and it is reasonable to expect other miners will follow suit.

Final Thoughts

To their credit, decentralists have done an excellent job at alerting the community to the centralization and security risks of raising the block size. Those in favor of large blocks would do well not to ignore them.

By the same token, I have observed decentralists tend to dismiss the importance of transaction capacity while insisting that Bitcoin can be used for settlements, perhaps not realizing that any meaningful settlement network would require a significant increase in transaction capacity.

The Lightning Network is typically raised to reassure the ecosystem that transaction capacity will be resolved by this and similar layer 2 networks and operators. That may indeed be so, and while I think there are some non-trivial technical challenges, I’m pretty excited to see it in action… if and when it’s ready and well-tested. And of course, it must pass the most important test: consumers.

Likewise, those in favor of large blocks should also remember that efficiency improvements to the blockchain, such as sharding and invertible bloom filter lookup tables, are not ready for deployment yet either.

Let’s try and listen to each other.

Dr. Washington Sanchez

Send corrections or comments to @drwasho

Special thanks to my proof readers:

Ian Grigg
Angel Leon
Rippler

Appendix

Bitcoin

About

A peer-to-peer electronic cash system (this is what Satoshi called it)

Sources

Transaction Capacity

~94.6 million transactions/year
~259,200 transactions/day
~3 tx/s

Block size: 1 MB

Theoretically Bitcoin can handle 7 tx/s; but in practice it is somewhere between 2–3 tx/s
~3 tx/s assuming an average tx size of 520 byte transactions (see 3rd source)
Multisignature transactions (the kind needed for the LN) are slightly larger