Blog article
See all stories »

We Need Consensus on the Best Approach to Consensus

As has been noted elsewhere, the Hyperledger project is experiencing spectacular adoption and growth as the combined benefits of a permission blockchain fabric grounded in open source and open governance is even more widely recognized.  Indeed membership has risen from 30 to more than 80 participating firms now including DTCC, JPMC and CLS.

One of the defining features of blockchain is the means by which consensus is established across the various nodes of a network in order to accept a specific transaction. There are a variety of ways in which this can be done, including the wonderfully named Practical Byzantine Fault Tolerance.

In the world of Hyperledger options exist for how consensus is achieved.  Today this is based on the premise that all transactions are replicated across all validating nodes and security/accessibility is managed by cryptography.  This approach may work well for many use cases.  However, for the wholesale markets – where transactions are in millions of dollars, in highly regulated environments – such an approach will not be acceptable to corporate and investment banks.  And possibly not to their regulators either.

This issue has been well recognised within the industry by institutions such as R3 with CORDA, but now there is also a solution to this issue in the Hyperledger world; the so called NextGen Consensus.

Born in IBM Research’s Zurich Lab, NextGen builds on the benefits of the Hyperledger approach, but defines new nodes on the network.

Specifically, it defines:

  • Submitter nodes – Which receive a transaction from a counterparty’s application, and subsequently submit it to…
  • Endorser nodes – Which perform the validation of the transaction. Critically, these can be set up to be just the participants of a transaction (i.e. the trading counterparts and possibly the regulator or market infrastructure entity). Being customisable the endorser nodes solve the main issue described above, whereby a trade is never present (encrypted or otherwise) on the node of an institution not a party to that trade. Once the transaction is agreed to by the Endorser nodes, hash of the transaction is submitted to…
  • Consensus nodes – Of course a blockchain environment still needs to achieve consensus across participants. This is achieved in the NextGen approach by having a much larger number of consensus nodes that achieve consensus based on a hash, or cryptographic thumbprint of the trade. In this way, consensus is achieved in the network without the need for all nodes to have access to the trade.

Besides the benefits of security/privacy, the NextGen consensus approach does not need all nodes to validate all transactions, so fewer system resources will be required and scalability should be improved.  For markets use cases with demanding volume requirements, this may also prove to be critical.

It’s probably no understatement to say that this development greatly increases the attractiveness of Hyperledger to wholesale financial institutions.  I’m aware of several that now plan to implement major blockchain projects solely because of the development of NextGen consensus. Many in the industry are now looking forward to the integration of NextGen consensus into Hyperledger in the coming months.

– To see the details of the NextGen consensus design, take a look at the Hyperledger GitHub
https://github.com/hyperledger/fabric/blob/master/proposals/r1/Next-Consensus-Architecture-Proposal.md

– To contribute to the thinking – join the Hyperledger community
https://github.com/hyperledger/hyperledger/wiki

– To join the discussion – meet IBM’s blockchain gurus at SIBOS in Geneva
http://www-935.ibm.com/industries/banking/events/sibos-overview.html

 

3890

Comments: (1)

A Finextra member
A Finextra member 28 September, 2016, 06:171 like 1 like

From the write up I understand that ' consensus' nodes form a separate group. If for example there is data tampering at the submitter's or endorser's level, it will not be known till such time the consensus nodes confirm the hack. This appears more hierarchial than peer to peer. However, I agree on the point that there is minimisation in the use of system resources. .