Financial analytics company accelerates fraud processing 10X with GridGain

A financial analytics company originally replaced their legacy database stack (Oracle DB, RedHat Elastic, MongoDB) with open-source Apache Ignite to gain speed and scalability. GridGain optimized this deployment, achieving a 10x performance boost. However, a major client's demand for enterprise features—like high availability and stable snapshot backups—required them to switch from open-source Apache Ignite to commercial GridGain Ultimate Edition (GGUE). Today, their active-active data center configuration delivers 10-20 ms machine learning model execution at a minimum of 25,000 transactions per second with near-zero downtime.

Meeting Enterprise SLAs

• Critical Failure Point: The open-source Apache Ignite deployment could not support the enterprise demands of the financial analytics company's largest client—a top 20 global bank under regulatory pressure.
• Latency Spikes: Enabling the critical Snapshot feature for backups caused the P99.9 latency to spike from under 40ms to over 120ms, violating the client’s strict fraud detection SLAs.
• Mandatory Enterprise Features: The client mandated several capabilities not available in the open-source version:
  • Cluster-to-Cluster Data Center Replication (DCR)
  • Snapshot Backups that adhere to the P99 latency SLA (< 40ms)
  • Full Active-Active High Availability and Disaster Recovery
  • Built-in Security features to replace a custom-developed plug-in and reduce operational risk

GridGain Ultimate Edition

• Initial 10x Speedup: GridGain optimized the Apache Ignite implementation to achieve a 10x acceleration in transaction processing, moving the financial analytics company away from their slow legacy database stack.
• Final Upgrade: The company upgraded to GridGain Ultimate Edition (GGUE) and deployed GridGain Control Center.
• Rapid Delivery: Within months, GGUE delivered the required enterprise features—high security, stable backup/recovery (Snapshot), high availability, and real-time monitoring—allowing the company to onboard the critical client before their compliance deadline.

Performance and Stability Delivered

• Performance Metrics Achieved:
  • 10x Reduction in Overall Latency: Mean latencies of 20ms and P99 latencies consistently below 40ms.
  • SLA Adherence Under Load: The stable, enterprise-grade Snapshot feature eliminated latency spikes, achieving < 40ms P99.9 latency even during backups (down from 120+ ms).
  • High Throughput: Guaranteed minimum of 25,000 TPS for real-time ML model execution.
• Enterprise Capabilities & Impact:
  • Successfully implemented Data Center Replication (DCR), Security, and Snapshots.
  • The company's fraud management solution now supports six large banks and saves customers millions in regulatory fines and lost income from fraud.