One of the major arguments made by those against on-chain scaling solutions is that increasing block sizes leads to a reduction in the number of full-nodes, making the network more centralized.
According to a paper on block size increase^[1] by the BitFury group published in August 2015, the following factors contribute to limiting the number of full-nodes in the Bitcoin network:
Blockchain storage,
Average bandwidth,
Traffic per day, and
RAM usage
Comparision of 1MB vs. 8MB blocksize*
* ^(Assuming full blocks)
| Charecteristc | 1 MB | 8 MB |
|---|---|---|
| Storage per year (in GB) | 51 | 411 |
| Average bandwidth (in kB/s) | 148 | 1148 |
| Daily traffic (in GB) | 12.4 | 99.2 |
| RAM usage (in GB) | 4 | 32 |
They used a Steam hardware & software survey^[2] to estimate how much of a reduction in the number of full-nodes can be expected if blocksizes jumped from 1MB to 8MB and, according to the paper, it was a 90% reduction.
However, blocksizes wouldn't just jump from 1MB to 8MB immediately, the increase would be gradual. It took Bitcoin 8 years to fill a 1MB block size limit, it would take a few years at least before 8MB blocks start becoming full.
So, now, the question is how long will it be before Bitcoin can scale from 1MB to 8MB size blocks with zero full-node loss?
This can be answered by using some data from the paper, the Steam survey, Nielsen's Law^[3], and Moore's Law^[4].
Conidering each of the points separately:
Blockchain storage
It will take and additional 411 GB storage per year to store the blockchain. Accoring to the Steam survey, a majority of users already have total storage greater than 1 TB, which means that most average users have the capacity to run a full-node with full 8MB blocks for at least an additional 1.5 years before they have to get extra storage.
If we follow Moore's Law, the average user is going to have 2TB storage in 1.5 years anyway. And the average user can be expected to have 10TB storage in 5 years time and 110TB in 10 years time. Storage is very very unlikely to be an issue with 8MB blocks.
Anyway, for zero full-node loss, following Moore's law, a person who can store 51 GB per year now can store 411 GB per year with no additional cost in 4.5 years. As long as the increase from 1MB to 8MB takes takes less than 4.5 years, there should be no reduction in full-nodes due to storage costs.
Average bandwith
How long does it take for average bandwidth to go from 148 kB/s (for 1 MB blocks) to 1148 kB/s (for 8 MB blocks)?
According to Nielsen's Law, that should take about 5 years.
However, one should note that the global average bandwith is already around 1 MB/s. Which means that most users can already run a 8MB full-node with all full blocks now without bandwidth being a constraint.
Traffic per day
How long does it take for us to go from 12.4 GB dailly (for 1 MB blocks) to 99.6 GB daily (for 8 MB blocks)
Following Nielesen's Law, again 5 years. A person who can download 12.4 GB daily at present can download 99.6 GB per day in 5 years with no incease in cost.
RAM usage
The average system RAM according to the Steam survey is 8 GB presently. For 32 GB RAMs to be common, it should take about 3 years, following Moore's Law.
Conclusion
The hardware required to run a full-node with full 8MB blocks will cost the same in 5 years time as it does to run the hardware required to run a full-node with full 1MB blocks now.
Or, as long as we don't start seeing full 8MB blocks before 5 years, the increase in blocksize limit should result in no significant loss in the number of full-nodes in Moore's Law and Nielsen's Law holds true.
The network traffic per day would be the major constraint that may reduce the number of total nodes, and not storage or RAM usage.
Sources:
[1] Block Size Incease, BitFury Group, http://diyhpl.us/%7Ebryan/papers2/bitcoin/bitfury-report-on-block-size-increase.pdf
[2] Steam Hardware & Software Survey: October 2017, http://store.steampowered.com/hwsurvey/?platform=pc
[3] Nielsen's Law of Internet Bandwidth, Jakob Nielsen, https://www.nngroup.com/articles/law-of-bandwidth/
[4] Moore's Law, https://en.wikipedia.org/wiki/Moore's_law
PS: Please feel free to correct me if there are any errors in the data presented or if I have made any incorrect assumptions.:)
Not nevessarily. Who says the growth will be linear based on the average of all that time?
It took Bitcoin several years to reach 1000$, how long did it took for the next 1000$?
logarithmic growth could make an 8MB block full in the next year, especially since there is already a catch up needed as visible in the high fees.