WHY THIS MATTERS IN BRIEF
Mining blockchain uses up an extraordinary amount of electricity, equivalent to an entire country, so companies are trying to fix the issue.
Proof of Work (PoW), the common consensus mechanism that secures Bitcoin and numerous other cryptocurrency blockchains, has given the technology a reputation for hogging energy. And hogging energy is an understatement. Indeed, a commonly advanced argument is that an army of specialized computers all racing to solve some arbitrary math problem can wind up using as much electricity as a country.
However, while university’s like MIT are looking for new ways to solve the issue, and Microsoft, with its minerless Ethereum blockchain, and IOTA with their algorithmic based blockchain, are leading the way, scientists from IBM claim to have found a new way to “reshape and combine blockchain architectures,” including PoW, to arrive at what they call in a paper a “sweet spot for energy efficiency, scalability and security.”
Announced the other week their discovery stems from applying PoW to a very different use case, the Internet of Things (IoT) arena that would run blockchain nodes inside the connected devices.
The problem they faced is that, unlike specialized PoW mining hardware for cryptocurrencies such as ASICs and GPUs, IoT devices vary widely in their computational power and energy resources. After all, IoT is a broad category that includes everything from pocket-sized temperature sensors to internet-connected automobiles.
As such, all or some of the devices in an IoT network might not be able to solve very complex PoW puzzles. Hence the impetus to make PoW energy-efficient, according to the IBM researchers’ paper.
“Efficiency in IoT can be defined as an optimal utilization of hardware resources and energy. Therefore, in order to achieve that, the IoT devices on the blockchain should optimally utilize resources and energy to maintain and progress the blockchain,” the researchers said in their paper.
Their proposed solution takes advantage of the fact that not all nodes on a network have to engage in mining. Many dedicated bitcoin users, for example, simply run full nodes to check miners’ work and keep them honest.
Working on a testnet, or simulated blockchain environment, the IBM researchers have been dividing the nodes into small groups of 250 to 1,000 and then allowing an Artificial Intelligence (AI) algorithm to decide what proportion of each group should be doing the mining work, depending on the amount of power used by each node and the security required. This, they say, “gains optimum results in terms of conserving power while preserving security.”
“At the moment we look at blockchains like totally flat peer-to-peer systems, in which all the nodes have to do the same things, compete against each other to get that mining reward, for instance,” said Dr. Emanuele Ragnoli, technical lead at the IBM Research in Ireland. “But you don’t need everybody to do the same type of job.”
Ragnoli said he wanted to create a “layered ecosystem” in which different peers can do different things, thanks to clever algorithms which cluster nodes according to their capability and assign specific duties to them.
“Some of the nodes do the full PoW, like you have in Bitcoin,” said Ragnoli. “They do that because of the analytics behind the blockchain, which can actually see whether a device can do that kind of job, and place that device accordingly into a cluster of other devices that will be assigned a certain type of consensus.”
The “sub-blockchains” maintained by these node groups are then connected using interoperability technologies such as Cosmos and Polkadot. In a nod to this patchwork, the IBM Research team has dubbed its lab project “Hybrid IoT Blockchain.”
Stepping back though the IBM Research work is part of a broader push to create a future Machine to Machine economy, where devices would have their own blockchain wallets and trade with each other, for example, picture one self-driving car paying another for the right of way… and so on.
But Ragnoli is realistic about the scale of the IoT challenge for blockchains, saying this world is still a “huge set of leaps” away. Attempting to take on a bite-sized chunk, his team investigated how a Machine to Machine ecosystem might work in an industrial setting, connecting cutting edge manufacturing activities between a number of factories in the Netherlands – IBM wouldn’t identify the enterprises involved but said there is a consortium on the horizon.
“Nowadays in Industry 4.0, or manufacturing, you have many different factories that collaborate with each other to create a single product,” Ragnoli said. “So you have sensors, machines, even algorithms and analytics operating in the different factories, and inside the same factory, that the need to interoperate with each other.”
From linking these factory devices together with the hybrid model, IBM found that arranging nodes in clusters of about 250, with 7% of those sub-blockchains doing PoW, achieved the best in terms of scaling, without sacrificing the hard-won security associated with PoW.
“We are taking common consensus algorithms like PoW, the vision of Cosmos, etc., and we are altering ways of putting them together. The way we are designing this is like small Lego blocks, driven by the AI layer,” said Ragnoli.
The IBM Research project is notable because it suggests that the deterministic requirements of blockchains can be combined with the black box of AI, something they’ve talked about before but in the healthcare setting, allowing machine learning algorithms to alter the shape of blockchains to adapt to power or latency limitations, without compromising security, and as such, this would seem to open the door to a whole new design space.
“Why not augment the blockchain with analytics and AI algorithms that can actually shape the blockchain in a way that it helps it to overcome some of the limitations that are out there now?” said Ragnoli.
In the case of IoT, the way this works is the AI receives as an input the IoT devices that are on the system and the available resources of those devices. It also assesses the overall security requirements of the system and decides which and how many devices are mining, the PoW difficulty, the block generation rate, the block size and tries to balance between required security and scalability. Therefore, IoT devices can still perform their application-specific tasks, such as data processing, and concurrently continue to mine blocks.
So how could this work impact the world of cryptocurrencies? Merely saying that PoW just needs to be better organized is like saying the free market could be more efficient.
Ragnoli said there could be a possibility to alter the way trading systems work in a dynamic way with different currencies, adding, “I’ve not gone as deep as actually altering the cryptographic consensus within – although that is actually a very interesting direction to explore.”