UTXO appliance for Bitcoin Cash

Terab is an open source software project, but the end-game deliverable is not merely a piece of software but rather a computer appliance that happens to be a build combining an open hardware recipe with some open source software. The distinction is subtle but important. For end consumers, running a piece of software on their machine is mostly a binary thing: either it works or it doesn’t. From a business perspective, the situation is different however - and it just so happens that Bitcoin miners are companies. The whole mining ecosystem already “went professional” half a decade ago.

For a business, the cost of a software solution is best reflected through the TCO, the total cost of ownership. No piece of software can truly function in full autonomy, Skynet is not yet there to replace faulty hard-drives. A degree of supervision is always needed, and paid for by the company, one way or another. If your piece of software can live in the cloud, then economies of scale can bring the TCO of your software to almost nothing. However, miners are not going to mutualize their blockchain backends through a third party cloud (1). There are too many risks involved and not enough upside.

(1) A “cloudified” blockchain management is exactly what you get when you join a mining pool. However, it’s not miners who join mining pools, but rather “hashing devices”. I am reserving the term “miner” for entities who routinely produce blocks.

Also, for the miner, a slow block validation means an increased risk of seeing its newly minted block getting orphaned by the network, if another miner can get ahead while the block validation is still in-progress within the mining node. Thus, a predictably good UTXO performance is also a business requirement for miners, as extra latency adversely impacts the miner’s capacity to successfully claim blocks for itself.

Thus, Terab should seek to deliver a solution that minimizes the TCO for miners scaling the blockchain. Our vision is an appliance that would act as a plug-and-play solution for the UTXO dataset. This first version of this appliance is expected to take the form of a rackable server (maybe a 3U height). The setup of the Terab appliance would require little more than plugging a power cable and a network cable (maybe doing it twice for redundancy), and configuring the microservice credentials.

For most use cases, the primary downside of computer appliances is their rigidity. However, from the Terab perspective, we do not expect much radical novelty to happen with Bitcoin Cash, except a, potentially, very large growth of the blocks. The vast majority of protocol evolutions, such as newer opcodes, have no impact on Terab anyway. Thus, we suspect that this downside will be amply compensated by the upsides associated with an appliance.

The capacity limits of a given Terab appliance will be known in advance, so that the actual expiration date of the appliance itself is known weeks, if not months, in advance also (2). Keeping the failure modes of the Terab appliance as predictable as possible should go a long way into lowering the TCO for miners.

(2) While the appliance capacity is known in advance, the volume of transactions depends on future market conditions, which can only be forecasted. Yet, as block sizes are capped, it will always possible to plan for the worst case scenario, assuming that future blocks will all be full.

Then the Terab appliance, being associated to a rigid hardware design, would also deliver a predictable latency to validate blocks. Hopefully, Terab will deliver a predictably good latency, but not having to cope with plenty of other moving pieces, which compete for the same computing resources, should help to deliver this performance.

Packaging Terab as an appliance also has security upsides. Indeed, the attack surface can be dramatically reduced compared to generic setups. For example, Meltdown and Spectre two recent, clever and extremely far-reaching approaches to generate software vulnerabilities can largely be mitigated, by not allowing anything to run on the appliance but the Terab software itself. Again, tight security by design should go a long way into lowering the TCO for miners.