Indian digital payment solutions company Mindgate Solutions is moving its real-time payments system, which sits on a traditional relational database system (RDBMS), to an open-source distributed SQL database in partnership with Yugabyte.
Mindgate Solutions’ vice president for product engineering and innovation, Nikhil Mahawar told a recent Distributed SQL Summit Asia 2023 that the company has been using traditional RDBMS for the last eight years in its applications.
With the rapid growth of Unified Payments Interface (UPI) real-time payments in India, it was becoming difficult to deliver horizontal scale and enable active-active setup for the critical applications, Mahawar said.
UPI is a instant real-time payment system developed by the National Payments Corporation of India (NPCI) which can also be used in other countries such as Singapore, Vietnam, Malaysia, Thailand, Philippines, Cambodia, Hong Kong, Taiwan, South Korea, and Japan.
Founded in 2008, Mindgate has an office in Singapore and serves over 50 global customers in Southeast Asia and the Middle East.
Mahawar said the firm processes close to 3 billion digital payments monthly and was facing both real-time and operational challenges.
During peak volume periods, a large pileup of active sessions used to occur which often went up to 5,000 instances on a real-time basis, he said.
Mahawar said Mindgate selected Yugabyte as its inbuilt sharding mechanism helps distribute data across multiple databases, which can then be stored on multiple machines.
This splits larger datasets into smaller chunks which is stored in multiple data nodes, increasing the total storage capacity of the system, thus avoiding piling up transactions on a single data table, he added.
He noted that manual intervention during backups and archival causes operational issues. This led Mindgate to decide on migration strategies with Yugabyte.
Low latency reads
Mahawar said prior to the shift to the open-source distributed SQL database, during a high-demand period, the response time used to become 200 to 300 milliseconds, “whereas a good response time should be under 50 milliseconds”. This caused the system to choke applications, impacting high throughput, he said.
Yugabyte provided low-latency reads, high-throughput writes, and the sync replication between clusters across data centres across regions without any latency, he added.
Mindgate’s current setup gave a recovery time objective (RTO) of about 30 to 60 minutes during the business continuity planning (BCP) drills, he said.
With Yugabyte’s raft consensus algorithm, the data is replicated between multi-sites. In case the source replication server becomes unavailable, the replication of data helps to support disaster recovery and data availability, he said.
The database supports fully distributed ACID (Atomicity, Consistency, Isolation, and Durability) transactions across multiple rows, multiple shards, and multiple nodes.
It ensures database transaction validity, in the event of system crashes, power failures, and other errors, Mahawar said.
The core architecture of Yugabyte is linearly scalable, and it has an automatic failover and native repair mechanism. There is no single point of failure, he added.
It provides support for SQL and NoSQL in a single cluster.
Mahwar noted that Yugabyte gives the company cloud agility, with security. This has helped the team migrate the on-premises setup to the cloud.
Pointing out the advantage of the Change Data Capture (CDC) feature in Yugabyte, Mahawar said the team can sync up with any of the data warehouse or data lake systems given by the customer banks.
CDC provides technology to ensure any changes in data due to operations such as inserts, updates, and deletions are identified, captured, and automatically applied to another data repository instance, or made available for consumption by applications and other tools, he said.
