What is the difference between a CLS blockchain and a CLS blockchain?

People at big banks and financial institutions don't talk too much about the actual integration of blockchain and distributed ledger technology (DLT).

However, tricky topics like integration and production readiness are front and centre at CLS, the world's largest multicurrency cash settlement system. CLS began doing DLT proofs of concept with its members back in 2015.

Today it is part of the Hyperledger Project, and has also made a public commitment with its long-time partner IBM to deliver a DLT-based architecture using Hyperledger Fabric, during the first half of 2018. This will form the underlying platform for CLSNet, a payment-netting service for 24 currencies. Participants will also be able to submit instructions over existing SWIFT-based channels.

CLS is also working with R3, the consortium of large global financial institutions which is focused specifically on financial applications of shared ledgers.

Tom Zschach, chief information officer, CLS, said: "The important thing about CLSNet is it doesn't have anything to do with settlement or funding. It doesn't touch our central bank accounts, it doesn't access RTGS gateways. What we are doing here is introducing the technology as more of a foundational step than a transformational step.

"Distributed ledger technology is still at an immature stage, so it doesn't make sense to apply it for the first time to the most critical processes. We don't think that's prudent and it doesn't fit our risk appetite," he said.

Zschach was clear that this will be a proper business service that adds real value for CLS members. The new netting-service will take trades that are not settled on CLS and net payments on a bilateral basis and then return payment schedules to the CLSNet participants. They then go ahead and process the payments in a standardised manner.

IBM and Hyperledger v.1.0

CLS is working with IBM to ensure that the DLT element of the CLSNet platform meets the necessary requirements in order to deliver a resilient and secure service. The technology should have no problem scaling to CLS's requirements, states IBM. The new version of Hyperledger Fabric (version 1.0) is designed with scalability in mind.

Keith Bear, vice president, financial markets, IBM, said: "Volume is an important consideration but the technology is perfectly capable of supporting the levels we are talking about in the case of CLS, also with Northern Trust and the other projects that we have going on at the moment."

Hyperledger v.1.0 has a new consensus model that means it doesn't have to reflect every transaction on every node. "There is a policy definition, which says which nodes have which transactions, implemented within individual 'channels', and this reduces the complexity of the consensus model and increases scalability as well," said Bear.

"Essentially, there is an endorser mechanism and a consensus mechanism, and the endorsers have access to the actual trade – if we are talking in a trading environment – but the consentors do not. The endorsers could be two banks that have a transaction and maybe the regulator, however you want to set it up. The other consenting nodes don't have access to the transaction in encrypted form or otherwise," he said.

In this respect, Hyperledger Fabric version 1 has reached the same conclusions as Corda, the DLT system built by R3. Bear pointed out that R3 has made an intention to provide Corda into the Hyperledger Project, and acknowledged that IBM is "collaborating with them elsewhere, DTCC being the example. They are subcontracted to us as part of the Trade Information Warehouse programme delivery at DTCC".

Charley Cooper, managing director at R3 said this is a really interesting time in the industry, where on any given day the same companies can be one side of the house working as partners, or on the other side of the house competing.

R3 Corda distinction

Cooper wanted to emphasise the distinction between the requirements that R3 and IBM have used to shape their platforms. He pointed to the fact that Corda is written for Java, the language of banks, uses standard financial message queues to distribute data only where needed, and stores data using mature enterprise relational databases, while Fabric is using a noSQL data store, for example.

Cooper said: "The folks at IBM Fabric who donated it to Hyperledger are looking to solve a whole set of problems for a much broader set of industries. So healthcare, auto financing, supply chain management, finance etc, whereas Corda was built specifically for financial institutions and built with the financial institutions.

"I think what CLS is doing is the right thing, meaning they are working with multiple different providers; they are involved in Hyperledger work with Fabric, they are involved with the R3 consortium, and working on projects that go beyond that."

Clearly there is some healthy competition between R3 and IBM and also a bit of history with R3 and Hyperledger. But there is also a fruitful meeting of minds on the DTCC project.

DTCC is doing the right thing by focusing on a very small asset class first within its trade information warehouse to test run the technology and sure it's fit for purpose before it tries to replace the whole system, said Cooper.

"Because this technology is still in early stages and key design decisions are being made, rather than go all in with one vendor, DTCC decided to go with a group of us.

"This is the really interesting thing: sometimes we are competitors; sometimes we are working together. Here we have come together. We have at times differences of opinion and we challenge each other. But certainly in the context of the project, this actually makes us better.

"So Axoni is coming in with the actual software, it's their software design; IBM is looking at this really as sort of the systems integrator and project management piece. What we are bringing to the table is the collaborative mindshare of the world's financial institutions. Ultimately they are the clients of DTCC, and if you don't design the solution up front to satisfy their needs, the whole experiment will fail," said Cooper.

Pushing complexity onto the network

In the case of CLS, Zschach has set out key requirements from integrating distributed ledger technology, one of which is to increase resiliency. CLS will use DLT to discover how much responsibility can be pushed out to the nodes on the network.

"From a resiliency point of view, if all of the nodes could trigger a netting rent, for example, or process their portion of the netting run, the system could keep operating, even in the event of a failure or temporary interruption," said Zschach.

"However, a participant wouldn't be able to come in and say, 'I want to create a new cut-off time for a currency', for example. We run the service; we have the rulebook; we have to be able to control and manage everything."

Once a network is capable of handling parts of the process there is potential to hook in data providers, for example, or a KYC/AML Office of Foreign Assets Control scanner. "It presents an example of how you can continue to build on top of the network to the benefit of everybody that's involved," said Zschach.

It's important to think outside of the PoC box and consider questions you will face in a live environment, said Zschach. In the future a big bank might be looking to participate in more than one blockchain: trade finance on the one hand, and another for repo at DTCC, for example.

"How are we going to connect these ledgers; how secure will that be; how do we handle certificates; how do we hook into participant bank's hardware security module? Do you want different ledgers, different security standards? All of these questions are extremely important in a live environment," he said.

"Perhaps standardisation on something like a DLT cloud or similar could be the answer. Otherwise we risk taking the complexity out of financial institutions and simply throwing it into the network."