Bitcoin Network Capacity Analysis: Data Propagation

Bitcoin Data Dissemination

The timely propagation of Bitcoin data in the network is an essential basic function of this ecosystem. The propagation speed can be measured by connecting to a large number of network nodes and collecting and storing them in a real-time database. TradeBlock has a large cross-regional Bitcoin network data architecture that can view and record every message broadcast to the network, including those that are not widely broadcast, thereby providing an accurate analysis of the network.

Average propagation speed

The above figure shows the average block propagation speed in 2015. The average block propagation speed is measured by tracking the block data broadcast by each node on the Bitcoin network and calculating the time interval between the first broadcast and all subsequent broadcasts. This figure can show how the block size affects the data propagation speed and how it affects the incentives and behavior of miners.

Propagation speed and orphan race

When two or more miners spend about the same amount of time calculating a block, the Bitcoin network will simultaneously provide two (or more) potential options for calculating the next block. Since miners will focus their computing power on the chain with the most work (usually the longest chain with the most blocks), the speed of propagation is crucial in this case. Generally speaking, the block with more nodes has a greater probability of winning the orphan block competition, allowing the next block to be connected to it.

In addition, rejected blocks are called "orphan blocks" and are not accepted by subsequent blocks. The miner who processes this block cannot obtain the coinbase reward. The figure below shows the number of orphan blocks mined every day since mid-April. On average, 1.3 orphan blocks are generated every day, which means an orphan block rate of about 1%.

Daily orphan blocks

The following graph is an observation of orphan races over the last three months. Although there are many reasons for orphan blocks, it is obvious that the most typical reason for unsuccessful blocks is that they have fewer nodes connected (relative to the successful orphan races). Although there are about 6,000 active nodes in the Bitcoin network at a given time, the data shows that as long as there are 3,000 connected nodes, the probability of a particular block winning the orphan race will be 90%. Finally, the winning blocks in the orphan race seem to be usually the first to be broadcast to the network, which is also confirmed by the fact that the winning blocks in the graph mostly appear in the positive direction of the x-axis.

Orphan Block Race

Next, we examine the average size of blocks participating in orphan block races over time. The figure below shows that the average size of blocks participating in orphan block races is 20% larger (about 100kb) than blocks not participating in orphan block races, which may be because large blocks take longer to broadcast.

For marginal transactions in a block, each node must check with the block transaction list stored in its own memory pool to finally confirm the block. The result is that the more transactions there are in a block, the longer the data propagation time will be, and the more time it will take for each transaction to jump around the network. This means that other miners will spend more time hashing the previous block before the found block is broadcast.

Orphan blocks and non-orphan blocks

Propagation Speed ​​vs. Block Size and Miners

To demonstrate the importance of data transmission, we analyzed the correlation between propagation speed, block size, and miners. Our data was collected from April to June 2015. As shown in the figure below, when the block connects a critical 3,000 nodes, there is a direct relationship between block size and propagation time. For example, a 700kb block takes 17 seconds to propagate, while a 200kb block only takes 6 seconds.

Time required to broadcast to 3000 nodes

We also explored the relationship between miners and the average time when a block is broadcast to 3,000 nodes. As can be seen from the figure, the average time spent by most miners falls within the standard deviation of the mean between 9.7 seconds and 7.2 seconds, with the highest (21 seconds) being a Chinese mining pool BW Pool and the lowest (4 seconds) being a Polish mining company Polmine.

By marking the geographic location of each block's packaging miner, we can analyze the relationship between propagation time and region from a broader perspective, as shown in the figure below, which shows that regional differences will cause slight changes in propagation speed. Overall, European miners are slightly faster than the average speed, while the propagation speed of China and the United States is almost the same (when removing the highest value BW Pool).

The time measurement of Chinese mined blocks is only feasible if the Chinese GFW is not taken into account. If the GFW is taken into account, the true propagation time is unknown. This means that when the nodes that TradeBlock is connected to (less than three digits) monitor data broadcast from Chinese IP addresses, it is likely that it has been blocked by the GFW in a timely manner.

Data propagation time and miners

Data dissemination time and region

Implications for the Blockchain Expansion Debate

The above analysis has important implications for the ongoing debate over block size, as we have demonstrated that (1) block size is directly related to the time it takes to propagate through the network, and (2) blocks in orphan races are significantly larger than blocks in non-orphan races.

Based on the data propagation speed when the block size is 1MB, we can infer the propagation time required for larger blocks. As shown in the figure below, a best linear regression shows that it takes about 137 seconds for an 8MB block to propagate through the network to 3,000 nodes.

Prediction time

Although we have suggested in previous research that people should not expect to expand blocks to 8MB in a short period of time, because every extra minute it takes to propagate data means an increase in the probability of orphan block races, which in turn incentivizes miners to pack fewer transactions. Several proposals have been proposed to address this problem, including mining backbones, transaction cache updates, and reversible Bloom lookup tables, but before these proposals are implemented, the propagation of large blocks may bring risks to mining revenue.