Recently, there is a discussion in Ethereum Classic community on changing the currently-used mining algorithm, Ethash. It resulted in some controversial debate among the community members on whether the change is approriate, and what will be the criteria for choosing the new algorithm.

Still up till today, how mining algorithms are chosen, and how evaluations should be done remain largely unexplored, with only various small pieces of information clattered in the Internet. Below, we try to provide three basic categorization of selection criteria for chooing a mining algorithm. Hopefully, this can be a small step forward in improving our vocabulary when we talk about mining algorithm selection, and can provide some further ideas to the Ethereum Classic community for the ongoing discussion.

Soundness

In this section, we talk about safety of a mining algorithm. Two important properties must be satisfied in order for an algorithm to be considered a sound choice for mining:

  • Uniformity: A good mining algorithm should produce roughly uniform distribution for its output range. This is mostly due to how difficulty calculation works in Proof of Work – the probability of generating a specific output is calculated and then used to derive how much work should be put in producing the next block. If a mining algorithm does not have uniformity, then the difficulty calculated becomes unrealiable.
  • Pre-image resistance: This makes sure the algorithm is a one-way function. This ensures that miners cannot cheat when producing work.

The above two properties are shared with hashing algorithms. However, mining algorithms and hashing algorithms do not have the same strong safety requirements. Mining algorithms safety requirement is slightly relaxed, in that it’s nice to have, but not the end of the world, if second pre-image resistance and collision resistance is violated.

Deliberation

Public blockchains are permisionless. Everyone is free to join, and free to leave at any moment. Miners as well. Because of this, one thing for Proof of Work is that we must ensure mining is a deliberate action. In order for the honest majority assumption to be satisfied, miners must be aware that they are mining, and miners must be aware that they are actively spending their own resources.

Below we discuss two examples of the opposite of deliberation.

  • Botnets: Those are attacker-controlled computation resources, such as vulnerable laptops, servers, IoT devices, or even routers.
  • Browser-based mining: Those are usually WebAssembly based binary running when a user visits a particular website. The binary then execute in the background, without awareness of the user.

Those two scenarios are examples where the owners of the hardwares do not aware that they are participating in blockchain mining, and do not aware that they are spending resources. It is easy to see why in scenarios like this honest majority assumption can be easily violated.

Deliberation has been an issue that many mining algorithms have been trackling. Common methods includes:

  • Memory-hard mining algorithms: In those algorithms, to efficiently participate in mining, miners are required to allocate a considerable amount of memory. This makes botnets and browser-based mining impractical. Even if the attacker managed to initiate the mining process, the large amount of resources consumed can be easily noticed by the device owner.
  • ASICs with massive efficiency improvements: The current most notable example of this is Bitcoin. Currently, ASICs are around 4 million times more efficient in mining compared with a normal CPU. In this case, even if botnets or browser-based mining exist in large scale, it’s hard to compete with a single mining farm.

Availability

Availability refers to how easy or hard it is for miners to acquire hardwares used for mining. Two entirely opposite approaches have been adopted in Proof of Work blockchains for this issue:

  • ASIC resistance: Some mining algorithms aim to be ASIC resistance. The goal is to make mining on general-purpose hardwares, such as CPUs or GPUs, to always be the most efficient. In this, the availability property is reached because miners can easily aquire or already own general-purpose hardwares.
  • ASIC friendly: Other mining algorithms choose an entirely opposite approach. The goal is to encourage design and manufacture of ASICs.

Years of attempts from a large number of PoW blockchains have shown that the availability issue is not a easy problem. On the ASIC resistance side, ASICs emerges for algorithms originally for this purpose, such as Ethash and Cryptonight. On the ASIC friendly side, centralization of mining power and closed-door hardware design have shown many issues, for example, Siacoin was forced to fork away due to a closed-door hardware design from Bitmain, which led to centralization issues.

Previous: HINT Instructions

September 7, 2019

Towards Performance Optimization Without Hard Forks

Up: Classic in Orbit