future-proofed performance
Proven performance for low-latency, high-throughput applications - Alibaba has achieved a throughput of 472 million events per second with zero-error sub-second latency on commodity hardware.
Atomic settlement
We estimate the annual 2018 RTGS volume of ~25 billion payments can be processed in less than 4 hours on just 32 cores. This provides significant headroom for domestic volume growth, conversion of DNS flows to RTGS (if desired), and expansion of synchronisation services to RTGS systems globally.
highly resilient
Designed for fault-tolerant, highly available applications that run 24/7/365. Infrastructure and storage can be distributed globally, offering resilience to multiple site failures and regional outages, with self-healing, auto-scaling infrastructure.
Guaranteed correctness
High-throughput serializable ACID transactions (the highest class of transactional consistency) with scalable distributed two-phase commit semantics, highly available exactly-once processing, and event-time processing for accurate state computation on real-time data.
reduced liquidity & settlement risks
We believe a low-latency, high-throughput platform with atomic settlement will result in a higher velocity of payments and lead to better liquidity optimisation with an opportunity to further address settlement risks.
real-time data intelligence
Real-time processing and analytics are combined, allowing an integrated approach for historical, operational and predictive reporting. This creates a rich data platform for regulatory reporting, business intelligence and machine learning models.
operational focus
Highly controlled development and deployment allows whole-platform redeployment and a flexible, low-risk approach to managing infrastructure, data, and software changes—including blue-green testing and what-if scenario modelling. Fully automated managed stateful upgrades, migrations and rollback/recovery using global state snapshots.
modern architecture
Centralised distributed ledger architecture based on an immutable event log means processing is fully deterministic; all results are predictable and can be reproduced exactly. Unified model for real-time and batch computation allows incremental migration of batch-based legacy environments.