Response time of less than 1 second achieved for a Fin-tech company in Singapore
Client overview
A leading fin-tech company that provides stock information empowering people to make smart investments.
Key requirements
Scalability assessment and performance optimization of the web application.
Smooth functioning for a concurrent load of 20000+ users.
Technological stack
Experienced in 10 + performance testing tools
Our Solution
A well-structured approach was put into action/deployed by the team, providing the client with an exhaustive review of system vulnerabilities and recommendations to get rid of them.
A Performance Benchmarking exercise was conducted to gauge the system capacity. This helped evaluate whether the IT infrastructure could handle the surge in transaction volumes without compromising service levels.
With a thorough Architecture Review across the stack, three failure points surfaced that had majorly hampered system’s performance. The key factors were namely- slow DB queries, low server capacity, and blocked calls by an external server
The MySQL database was reviewed at the architecture, query, instance, and I/O level to identify performance improvement prospects.
The infrastructure was examined to identify bottlenecks in the CPU, storage, and network.
The application server performance was analyzed to identify optimizations via thread, memory, and connection pool management.
Client Benefits
The solution provided by Frugal testing bolstered the responsiveness of the application, with response time being less than a second for most of the transactions, thereby further helping the client achieve cost, time and effort saving.
MAJOR IMPROVEMENTS DERIVED FROM PERFORMANCE TUNING AND OPTIMIZATION EXCERSISE
CODE LEVEL DIAGNOSTICS
Web Transactions response time during Load Test
Before
After
Changes at code level improved the responsiveness of the, application, making it 3X faster. It helped identify the transactions that took comparatively more time to process, and also the reasons for the delay by deep-diving into the method in the code causing issue.
Service Metrics
Before
Server Instance CPU
After
Server Instance CPU and Network In -Out
Monitoring end-to-end infrastructure and server-level metrics such as garbage collection, heap memory utilization, CPU, memory, etc., helped identify the performance bottlenecks.
Improved load compability
Before
After
Implementing auto scaling up and down at the server level helped in
handling the peak user load effectively. It increased server capacity from 0.1 hits/sec to 12 hits/sec.
RESPONSE TIME
Before
After
Top 5 time Consuming external services
The major business transactions communicate with external services. The external server, after a certain number of requests, was barring any further requests of the AUT. Diverting the direct calls from an external server and storing the data in the database through a batch job helped decrease the average response time substantially to less than 2 seconds according to SLA.
RESPONSE CODE
Before
After
Helped in fixing various 4xx and 5xx response codes in the initial
iterations.
"Uncover the Power of Testing"
Case Studies Showcasing Success in Various Testing Approaches
