Many things have been predicted to happen when Bitcoin reaches its capacity. Now, as it happens, the mempool becomes one of the most intriguing Bitcoin stats to observe.
When Bitcoin blocks reach its size limit of 1MB, there is a prediction of doom and gloom. Most prominently Bitcoin developer Jeff Garzik forecasted in December 2015 the so-called “fee event” that should occur when blocks are at least 95 percent full over at least seven days. According to Garzik, this will trigger an economic change.
Now, more than one year later, Garzik’s conditions are still not entirely met. There has been no week in which the blocks have been at least 95 percent full. But the system reaches its capacity, with more days of full blocks and fewer days of blocks with empty bytes. It is very close to Garzik’s fee event.
This gives us a fantastic opportunity to watch live how things play out. Is it as bad as some have predicted? Or has it all been an exaggeration, and living with full blocks is not that bad?
First, it must be stated that the current degree of fullness does not affect the price negatively. Transactions need more time to be confirmed, and fees for transactions are rising rapidly. But the price of bitcoin is not going down because of this. If there is any correlation, the price is going up, as it crosses the $1,000 mark again this week.
The most interesting data under conditions of full blocks is the so-called mempool, which refers to the pool of unconfirmed transactions each Bitcoin node kept in memory until it is confirmed and written in the blockchain. Jochen Hoenicke provides an excellent display of the mempool, where his chart shows it as fee-based layers. Each time a block is found, the mempool is reduced by 1MB. Usually, the miners take the transactions with the highest fees; hence the mempool is reduced from the top.
At the time of writing around 30,000 transactions have been in the mempool taking almost 30MB in space (not included around 10MB of zero fee transactions). To bring the mempool down to zero the network needs around five hours of blockspace.
Often when the mempool filled up, it was said that this is not the result of organic demand, but of an orchestrated spam attack which aims to cause fear, uncertainty and doubt or propagate some consensus breaking alternative clients. And in most cases, irregularities in transaction patterns indicated this to be true.
The current stage of the mempool does not look entirely native too. If you zoom out, you see on Sunday, February 5, the growth of a hill of a relatively few, but enormous, transactions sent with extraordinarily low fees. This hill can be easily detected. It consists of less than 1,000 transactions with fees of five to 10 Satoshi which cover 18MB in the mempool. One thousand standard transactions would need less than 0.5MB.
At the same time, the UTXO set was reduced. This means the whole of the “unspent outputs,” which are all unspent coins connected to addresses. It contains every bitcoin in existence, the whole of the digital coins.
The correlation of strange transaction patterns and the reduction of the mempool indicates that either there has been a sophisticated spam attack which operated with a huge number of transactions. Or it shows that some Bitcoin companies have begun to use days like Sunday, which usually see little traffic on the blockchain, to reduce its own UTXO set by cumulating a large number of dust outputs in fewer outputs. Such actions would be a rational strategy to reduce operational costs for every company receiving a large amount of smaller transactions.
If this is true, the company might be in trouble now, as around 10MB of these transactions are still in the mempool, and it is not very likely that they will be released soon.
For most other users the current state of the mempool is a nuisance with limited impact. Jochen Hoenicke’s visualization shows that in few moments there have been more than some MB of transactions with fees higher than 70 Satoshi each byte, which equals something like 35-40 cents for a typical transaction. Everybody who is willing to pay this amount has a high certainty that their transaction is confirmed the next one to three blocks.
Despite this, a look at a larger timeframe of Hoenicke’s chart yields a more concerning picture; the intervals between the mempool spike get alarmingly shorter.
During the first half of 2016, the mempool mostly looked like The Netherlands or Denmark; very very flat, only with minor hills in some exceptional situations (like the combination of bad luck of miners and a tough difficulty adjustment). Toward the end of 2016, the pool started to look more like Germany or Spain, a mix of vast flatlands and some hills. As we reach February 2017, the picture is reminiscent of a country like Switzerland; a region of mountains in which flatlands ceased to exist. Ever since February 1, the mempool was rarely smaller than 10MB, and every increase builds on top of this plateau.
If this is the beginning of a fee event with massively rising fees and an ever-growing mempool, or if the pool will clear over time, like it always did in the past, remains to be seen. But there is no doubt that Bitcoin enters a new territory for which no land map exists.