Javier Castro - Fotolia
The blockchain energy consumption problem is an issue for supply chain applications. However, it is not as significant as recent reports about the blockchain energy consumption problem for currencies like bitcoin may indicate.
Modern blockchain applications being built for supply chains consume less than one ten-thousandth of the power required in currency applications. However, blockchain and distributed ledger technology (DLT) applications for the supply chain consume less power than bitcoin but significantly more power than comparable supply chain applications built on top of traditional databases. Still, some experts believe this may be significantly less power, time and money than an enterprise might spend manually sorting out supply chain problems with humans making phone calls and sending emails.
It's clear there is a blockchain energy consumption problem for supply chains, but it is not a deal breaker. Rather, it is comparable to the level of waste in modern application development practices that allow companies to experiment with app features and business models faster than competitors. In the long run, the benefits of using blockchain in supply chains to improve transparency, auditability and to reduce manual processes could outweigh any problems caused by blockchain's energy overhead. There are more significant blockchain adoption problems in supply chains for enterprises to consider.
Energy consumption relates to computation
Concerns about blockchain energy consumption problems for supply chains most likely originated in recent reports suggesting that the process of creating digital currencies like bitcoin now consumes about 1% of the world's power supply. The vast majority of this energy use comes from the computations required to create currencies, called proof-of-work algorithms. These algorithms were designed to be inefficient to limit the number of bitcoins.
"Computing the ratios of energy consumption requires detailed models. However, one can estimate it based on the amount of computation necessary," said Monish Darda, CTO and co-founder of Icertis, an enterprise contract management platform. In supply chain applications, a blockchain serves the same function as other types of data stores like structured query language (SQL), key-value pair (KVP) and other NoSQL databases. A centralized database involves around 10,000 computational steps, whereas bitcoin currency mining requires billions of computational steps using brute-force search algorithms. You can comfortably estimate eight to nine orders of magnitude difference in terms of computation, Darda said.
In contrast, supply chain applications use far more computationally efficient algorithms for logging the status of goods throughout the supply chain. One big factor is that proponents of supply chain networks are architecting blockchain applications with permissioned servers that trust each other.
Martha Bennett, a principal analyst at Forrester, said, "While not all parties in a DLT-based network have direct relationships with each other, and hence may not trust each other directly, there is an overall trust foundation inherent in a permissioned network, provided it has been architected and implemented to the required enterprise standard, with an appropriately designed and curated governance model."
None of the enterprise-focused platforms, such as Quorum; Hyperledger Fabric and Sawtooth; R3 Corda; Digital Asset or MultiChain, rely on bitcoin's proof-of-work mechanism because it's too slow and expensive.
Bitcoin represents an attempt to replace trust mechanisms already baked into supply chain networks like personal relationships and legal contracts with algorithms. More efficient blockchain algorithms are possible when participants trust each other. "Enterprise and consortium blockchain solutions are permissioned and don't rely on mining, so typically use far less energy," said Bill Clabby, research analyst at Clabby Analytics.
The overhead when using blockchain for distributed supply chain applications could be estimated in terms of the additional computation required for each algorithm, which is one to two orders of magnitude more in terms of computation, and time, compared to a centralized database. Darda said this estimate could be affected by details like transaction record size and number of stages.
However, Darda stressed that it's hard to make a direct comparison in terms of computational efficiency. "What is obtained at the end is not merely a record of a transaction, but also additional things like distributed trust and reliability, which are not relevant for a simple, centralized database system. Comparing these two different strategies in terms of efficiency may not be possible."
One of the reasons is that traditional application and data management architectures for supply chain management were designs that focused on addressing needs within organizational boundaries from both a technology perspective and compliance perspective. "Establishing trust and transparency across multiple layers of suppliers remained a hard problem to solve," said Murali Karthik, vice president and global head of business development at Mindtree, a technology consulting and services company.
Blockchain transaction speed a factor in energy consumption problem
The biggest component of blockchain energy consumption is the number of computations required for a given operation. It takes more computations to store or retrieve a record from a blockchain compared to traditional data stores. More computations relate to the speed required to execute a given process in a supply chain transaction. This means that a blockchain application may be slower than a supply chain app running on a traditional database.
This speed penalty is not a direct correlation, since researchers are building better algorithms, and many of these algorithms can be run in parallel. For example, Fluree, an enterprise blockchain provider, claims it can append data to a blockchain at about the same speed or even faster than traditional data stores, but reads are about 10 to 100 times slower. "Keep in mind that over 90% of applications are read/query-heavy, not write-heavy, so this impact is moot for most," said Brian Platz, co-CEO at Fluree.
Bennett said that it can take up to 10 minutes to execute a bitcoin trade and an hour to confirm that it executed. By comparison, Walmart estimated that it can identify the source and destination of a batch of leafy greens involved in a food safety recall in 2.7 seconds on top of IBM's food trust blockchain. Although this may be slower than an app running on a traditional database, it is significantly faster than the three weeks that was required with Walmart's traditional manual process.
Bennett said there will always be an overhead associated with a DLT-based infrastructure compared with a centrally controlled distributed data store. This is in no way comparable to the waste of energy inherent in bitcoin. "But this overhead is why it's so important to be absolutely clear on why a DLT-based system is the best solution for a given use case," Bennett said.