Improving Core Banking Availability at Lower Cost: Oracle ADG Validation on Hyper-Converged Infrastructure

For core banking systems, high availability and data consistency are mission-critical, as any disruption directly impacts transaction processing, regulatory compliance, and customer trust. Therefore, many customers are using Oracle Active Data Guard (ADG) to provide real-time data synchronization, disaster recovery readiness, and workload offloading without compromising transactional integrity.

As business and data growth accelerated, a leading regional bank faced increasing difficulty maintaining required availability and performance while controlling infrastructure and operational costs. To address this challenge, the bank validated Oracle ADG standby databases on Arcfra Enterprise Cloud Platform (AECP) as a cost-efficient alternative to traditional architectures.

Production testing showed that AECP delivered significantly more stable and lower ADG latency than vSAN, with performance comparable to mid-to-high-end all-flash centralized storage during peak settlement workloads. The results confirm that AECP can reliably support core banking standby databases while reducing dependence on expensive centralized storage and lowering overall hardware and O&M costs.

Background

For core business systems such as deposits and loans, the regional bank utilizes Oracle databases in a Primary and ADG Standby configuration. The primary database is architected with physical servers and high-end all-flash centralized storage, while the standby runs on the VMware vSphere & vSAN platform.

To mitigate pressure on the core database amidst rapid growth in online lending requests, specific query workloads were offloaded to the standby database, and data processing tasks for reporting and submission were routed via the standby to the big data platform. However, operational usage revealed that the performance of the standby database was increasingly failing to meet escalating business demands:

• In batch processing scenarios, the standby resource pool’s I/O performance was insufficient to meet real-time replication requirements, resulting in an apply delay exceeding 30 minutes during daily settlement batches.
• The big data ETL extraction initiated immediately after the daily settlement batch exhibited a 30-minute data lag, resulting in inaccuracies in reports, regulatory submission data, and even front-end query data.

Concurrently, anticipating significant subscription cost increases following Broadcom’s acquisition of VMware, the bank planned to upgrade the standby databases with a more performant and cost-efficient vSAN alternative.

Initially, the user evaluated the solution of migrating the standby database to the existing all-flash centralized storage. Although this approach can fulfill I/O requirements, as the centralized storage already supported the primary core transaction database and other high-I/O standby databases, this might result in high expansion costs, alongside potential risks of performance interference.

To balance performance, cost, and reliability, the bank decided to assess the feasibility of replacing vSAN with an alternative hyper-converged infrastructure (HCI) solution for hosting standby databases. Having already deployed AECP (featuring AVE) in production to support MySQL slave databases for over 100 production systems — excluding the core accounting Oracle databases — the bank planned to validate the platform’s capability to host Oracle ADG databases.

Test Case

The test was conducted in the production environment using a comparative analysis approach, with standby databases running simultaneously on three clusters: vSAN, all-flash centralized storage, and AECP.

Cluster configurations were standardized, with a focus on monitoring ADG latency to ensure reliable real-time data replication. The test simulated real-world business scenarios, particularly during peak daily settlement batches, utilizing monitoring scripts to capture standby latency data in real-time.

Test Result

Monitoring over one month revealed that while the vSAN cluster largely met latency requirements, performance instability caused multiple severe spikes exceeding 900 seconds. Conversely, the all-flash centralized storage and AECP clusters demonstrated consistent stability, maintaining latency below 100 seconds during peak periods. With maximum latencies of 78 seconds and 85 seconds, respectively, their performance significantly outperformed the vSAN cluster.

These test results demonstrate that AECP satisfies the real-time performance requirements of core business operations when handling high I/O loads, particularly in real-time data replication scenarios. Looking ahead, the bank plans to gradually extend support for existing Oracle primary and standby databases by utilizing AECP and Arcfra Block Storage (ABS) tailored to specific use cases.

Key Outcomes

1. Validated the capability of running core banking standby databases like Oracle ADG on AECP.
2. Validated the capability of distributed storage to replace mid-to-high-end centralized storage, providing a new approach for banking users’ digital transformation in the era of all-cloud data centers.
3. Hosting core databases on AECP (HCI architecture) significantly reduces reliance on mid-to-high-end storage hardware in traditional solutions, while simultaneously lowering hardware procurement and O&M costs.

You may also be interested in:

Securities Low-Latency Trading System on Arcfra: Solution and Validation Test