Decentralization, Scalability, and Fault Tolerance of Bitcoin

We tend to get three different attributes mixed up causing endless confusion even among men of good will.

1. Throughput Scalability
2. Fault Tolerance
3. Decentralized Control

These are three different concepts.

Fault Tolerance

We are saying 101 highly-reliable, tested and proven, hand-picked parts dispersed across the globe and selected by the entire owner population is sufficient redundancy to achieve reliable fault tolerance. The only thing those parts can do is – do their job to spec. We can observe their performance and swap them out in ten seconds if they don’t perform to spec. So, really, they are just interchangeable slave machines. Producing the blocks is a mindless task.

Selecting which parts make up the machine is where the power lies.

Decentralized Control

The total decentralized population of the all owners participate in selecting the most reliable machines to run the network. Those 101 parts have no power over the owners. 101 dispersed redundant parts is a decentralization red herring! That’s not where control lies. Those 101 chosen nodes can be completely reconfigured or replaced by the fully decentralized participating owners in 10 seconds.

Throughput Scalability

Any of the 101 nodes can scale up by adding parallel machines, side chains, and a thousand inventions we haven’t dreamed of. When we get to a billion owners and need a thousand machines per node, we can do that. It will still be decentralized enough to ensure that the 101 (now bigger) parts that make up the distributed, fault-tolerant machine will perform their mindless slave jobs reliability – from positions scattered across 24 time zones.

Those million decentralized owners do not want to have to think about managing more than 101 redundant parts to their machine. 101 is plenty, maybe too many, for the average owner to keep track of how they are performing. Adding more parts reduces the degree to which each part can be vetted and therefore reduces the system’s reliability. Total reliability is a combination of node redundancy and node reliability via reputation-based vetting.

In BitShares, absolute control is fully decentralized down to the votes of every single atomic BTS satoshi. You can’t get more decentralized than that.

Bytemaster’s Conclusions

At the end of the day it is all about decentralization of influence. It doesn’t matter that everyone is theoretically able to enter the lottery to gain the honor to produce a block if the same people win the lottery the majority of the time. Economies of scale will always force systems toward centralization via the removal of unnecessary redundancy in order to compete in the free market. Absent a means of delegating block producing influence to another party, any proof-of-something system will tend to deny small players any influence.

Bitcoin & Nxt both support delegation of influence through mining pools and leased forging. Bitcoin and all of the alt-coins have proven that left to its own devices the market will realistically support no more than 10 pools. As the network scales block producers will be competing to lower fees. This inevitably means placing their full nodes in locations with cheap bandwidth and computers optimized for signature verification and database speeds. Users of the network will eventually start creating transactions with fees lower than would be profitable for any unspecialized block producer to process. It just isn’t worth the time, money, and effort for companies to run a block producing node at scale.

Nxt leasing and Bitcoin mining pools will ultimately result in fewer than a dozen organizations producing over 51% of all blocks within the confirmation window. I did a random sampling of Nxt block producers and in a matter of minutes identified 10 people who have collectively produced 30% of all blocks in Nxt history. At this rate I could likely find that 51% of blocks being produced by fewer than 40 people. There may be over 300 people participating, but the vast majority of blocks are produced by a small minority.

What we learn from this is that we must decentralize control with a different mechanism than we achieve fault tolerance. It is more important to have decentralized control over who produces blocks than to have no control over who produces blocks. If token holders don’t control who produces blocks, then centralized block producers will control who owns the tokens.