Multi-Region Dedicated Server Infrastructure: Reducing API Latency in Asia by More Than 3 times
About the project
A financial analytics platform collects market data from multiple sources, processes it, and makes the results available through APIs and a web application. The platform serves customers in both Europe and Asia. Its original infrastructure was concentrated almost entirely in Europe.
Challenge
The European architecture provided acceptable performance for local customers but resulted in noticeably higher latency for users in Asia. For some Asian users, round-trip latency to the API reached approximately 180 – 220 ms. The engineering team wanted to move latency-sensitive services closer to customers without migrating the entire infrastructure stack or managing multiple unrelated hosting providers.
The key requirements were:
- dedicated infrastructure in Europe and Asia;
- low-latency connectivity for API workloads;
- powerful CPU resources;
- NVMe storage;
- simple capacity expansion;
- one infrastructure provider across multiple regions.
Solution
The infrastructure was distributed between European and Asian Unihost locations. The primary database and data-processing environment remained in Europe, while latency-sensitive API services were deployed in Singapore.
Unihost solution used
- European infrastructure
MVS-09: Ryzen 9 9950X, DDR5 ECC, NVMe storage.
Used for databases, backend services, market-data processing, and internal infrastructure. - Asian infrastructure
VNF-05: Ryzen 9 9950X, DDR5 ECC, NVMe storage.
Used for API nodes, regional backend services, application caching, and other latency-sensitive workloads.
Relevant application data is replicated between regions. Requests from Asian customers can therefore be served from infrastructure geographically closer to them while core processing remains in Europe.
Results
After the Singapore infrastructure became operational, measured latency for users in the region decreased from approximately 180 – 220 ms to 40 – 70 ms, depending on network location. For some routes, this represented an improvement of more than three times. The multi-region dedicated architecture was also estimated to cost approximately 25% less than the comparable public-cloud deployment considered by the team. Operational complexity decreased as well because the company did not need separate infrastructure vendors for Europe and Asia.
What’s next
The architecture can be expanded with a third region, additional API nodes, dedicated backup infrastructure, and additional DDoS protection. Results at a glance: 180 – 220 ms – 40 – 70 ms latency, 3+ times improvement on selected routes, approximately 25% lower infrastructure cost, Europe and Asia through one provider.
