The following is from the initial discussion to reduce the validator set:
A brief look at the numbers
If we look at the current validator set, on an explorer such as NodesGuru , you will notice that only 39 validators represent two thirds of the voting power. Half of that voting power, one third, resides in the hands of only 11 validators. The majority of validators, 70 to be exact, represent less than one percent of the total voting power each. This means the network’s voting power is disproportionately distributed across the validators in an exponential, parabolic curve. But what does this mean?
Consequences and Implications
This means that the vast majority of validators are earning very little in terms of inflationary emissions. This undermines overall network security, as the primary purpose of validators is to secure the network via Proof of Stake protocol. If validator economic viability is undermined, it undermines their ability to provide the required infrastructure and services. It is not surprising to find that of the top twenty validators only two exceed a 5% commission. Conversely, twelve of the bottom twenty validators exceed this commission rate, likely as a result of increased economic pressure and reduced economic viability.
Consequently, larger validator sets require a higher inflation rate as overall network overhead is proportionally greater. This may further impact the network security through more pressure on token price. With a smaller validator set that makes meaningful network contributions, inflation can be lowered to the benefit of the overall network health.
A further implication is psychological. Not all validators can rely on large self-delegation and are completely dependent upon regular token holders to delegate to them. However, typical users do not delegate on a regular basis. They are very likely to select from amongst the largest validators to ensure their delegations remain in the active validator set. Combined with commission fee disparity this further undermines the viability of smaller validators. Comparatively small delegation changes may see them removed from the active validator set.
Practical Considerations
From both a security and economics perspective it would be better to have a smaller set of validators. Consider the following:
- A smaller set is easier to coordinate for technical updates, upgrades and the like, e.g. faster response to zero day vulnerabilities.;
- Greater economic viability for individual validators
- Less slices for a larger piece of overall delegations;
- Greater overall stability within the active validator set;
- Viable to offer smaller commission fees and remain competitive and sustainable;
- Increased validator stability and sustainability leads to better overall network security;
Finally, it provides the opportunity for better decentralization incentives… incentives for the support of smaller validators to progress towards a more linear (ideally horizontal) voting power distribution.
This does however represent some risks in the following areas:
- Potential technical issues resulting from a set reduction (such as around active delegations to validators removed from the set);
- Potential for transaction censorship as there will be less validators contending on block proposals;
- Potential for validator cartels to form which reduces effective decentralization (however, large disparity in voting power distribution also carries the same risk);
Additional Context
Nakamoto Coefficient
The purpose here is to improve decentralization by making sure voting power is more evenly distributed across a smaller set of validators… with the current set approximately 10% controls one third of the network and less than 40% two thirds of the network. Having 30 validators with an even distribution can significantly improve these ratios and result in a much better Nakamoto Coefficient.
At present only 7 validators control one third of Cosmos Hub and 25 validators control two thirds, and this in a set of 180 validators. More validators do not equate to better decentralization in a Proof of Stake network. However, better stake distribution across validators do.
We are also looking into inverse proportional staking with clamping as part of this recommendation. As stated we want to see a more even distribution of voting power across validators towards better decentralization. To achieve this we need to eliminate overt snowballing effects that simply allocate more delegations based on existing delegations…
Revisiting Validator Delegation Criteria
With this recommendation we are specifically interested in validators that do promote and remain active within our ecosystem, such as participating in governance, running relayers, etc. Large validators are just as likely to lose interest in promoting ecosystems that do not fall within the current “flavor of the day”; We want to ensure that validators that promote and remain active get to reap the long term benefits… simply increasing the number and amount of foundation delegations will only result in an increase of the effective inflation rate that further exacerbates the above mentioned issues pertaining to long term stability and sustainability.
Why 30?
As human beings we easily self organize in smaller groups, but once certain thresholds are reached these groups tend to fragment. This often occurs around the 12 mark. Thus, we need to ensure decentralization by selecting the number to be high enough so as to avoid easy syndication or consortia.
We selected 30 for the following reasons:
- It is still large enough to ensure effective decentralization;
- Mathematically it is highly composable (many common fractions and percentages are easily represented as whole numbers);
- It is very close to our current 2/3 voting power of 39 nodes;
