Australian exchange pauses project to move stocks to blockchain

Distributed ledger design slowed transactions, smart contracts were not so smart

The Australian Securities Exchange (ASX) has paused its multi-year effort to replace its core trading systems with a blockchain-powered platform, written off up to AU$255 million ($171m) of work, and been advised to reconsider whether distributed ledger technology has a role in the project.

"We have concluded that the path we were on will not meet ASX's and the market's high standards," admitted ASX chair Damian Roche in a filing [PDF]^*. "There are significant technology, governance and delivery challenges that must be addressed," he added.

Those challenges are outlined in a review of the project by Accenture. The consultancy found four underlying barriers that made the project unworkable:

• The current architecture "introduces higher latency" to the ASX;
• The system aims to conduct concurrent transactions, but doing so can cause contention when processing multiple in-flight transactions that target the same dataset;
• Batch processing was tried as a workaround, but did not fully address contention problems;
• Batching of transactions was constrained by practical limits in the Digital Asset Modelling Language Ledger (DAML) API and the size of messages that can be processed by the VMware Blockchain. Further work is needed to redesign and test around those limits, which Accenture warned "could be a hindrance to extensibility."

Some of the issues outlined above were rated as having the potential to create outages, or slow trades. Neither is acceptable for a securities exchange.

The project aimed to replace the ASX's existing application – called CHESS – with a system that used blockchain to record trades. The vision was to have market participants run their own blockchain nodes, with orders rippling out across copies of the distributed ledger. The project had the tricky requirement of replacing a 25-year-old application without disruption to market participants.

Accenture was called in to review the project after it blew deadline after deadline.

Those delays were mostly laid at the feet of an outfit called Digital Asset that champions Digital Asset Modelling Language (DAML).

The Accenture report finds that Digital Asset (DA) and the ASX did not work well together.

"Siloed execution and reporting between ASX and DA have resulted in misaligned views of status including delivery progress, risks, and issues," the report states.

"Management of vendor accountabilities is lacking, including inconsistent information obtained regarding the reporting and tracking of execution outcomes and quality-related metrics," the report adds, along with the observation of a "siloed client/vendor culture" that failed to drive the project towards shared business outcomes, which was "amplified by independent management structures, locations, and tools."

Project management was heavily criticized for lacking "a holistic, agreed, single view of status with adequate traceability of resources and estimation to the Draft Delivery Plan." That's consultant-speak for "no-one knew what was going on."

VMware's technology was considered out of scope for the review, but the document finds the ASX's application "primarily uses the ledger for resiliency which adds undue complexity to the solution e.g., the consensus contributes to the round-trip latency."

That latency problem was largely caused by DAML not executing smart contracts elegantly – which happened because DA and the ASX weren't on the same page.

In what may be a fatal blow to the whole project, the report also suggests distributed ledger technology (DLT) might not be the right choice. The document notes that the application design leaves the ASX as "the central source of truth and final arbiter of outcomes, minimizing many of the benefits of a DLT architecture."

Accenture therefore recommended the ASX "Revisit or refresh the DLT strategy to determine long-term use" – but also reconsider whether its business processes should be changed to make them more suitable to work as smart contracts.

The ASX has appointed a project director to revisit the solution design, establish new project governance arrangements, strengthen vendor management, and position the project "for the next delivery phase."

"This will include a process to select delivery partners to help address the capability gaps including those identified by the independent review."

The ASX announced its project in 2017. It was one of the first mainstream organizations to bet on blockchain for a mission-critical application.

Blockchain itself has not failed at the project. But implementing it to achieve the transaction volumes and speed required by a securities exchange has proven more difficult than anticipated.

As the Accenture report details, that difficulty looks to stem largely from the kind of misunderstandings that derail many software projects: requirements tied to legacy software that proved more complex than expected, plus difficulty putting an untested technology to work in a high-pressure environment.

Those challenges will not be unfamiliar to developers or project managers anywhere, many of whom may consider elements of this mess avoidable.

The ASX, meanwhile, says it has hardened its existing CHESS application and environment, which The Register understands was built using COBOL and runs on Intel's defunct Itanium CPUs. ®

^* The ASX is itself listed on the ASX, so it files documents just like any other listed entity.