Overview

This performance testing was conducted in three distinct phases to validate the stability, scalability, and cost-efficiency. The testing scope progressed from establishing single-tenant benchmarks to validating high-concurrency loads, and finally, evaluating a proposed hybrid database architecture.

The specific objectives for each phase were as follows:

Phase 1: Single Tenant Baseline

The initial phase established the performance baseline for the singe tenant(cs00000001_0002) operations on the segcon environment (DB r7g.4xlarge).

• Scope: Execution of standard user workflows—including Authentication, Circulation (50 checkouts/min), Search (200/min), and Inventory browsing—for a single tenant.

• Goal: Define baseline metrics with a target API p95 response time of <1s and Database CPU utilization under 20% to serve as a reference for scaling tests.

Phase 2: High-Concurrency Scalability (30 Tenants)

Upon establishing the baseline, the load was scaled to simulate a realistic multi-tenant environment.

• Scope: Testing on 30 concurrent tenants, simulating typical usage by multiplying the single-tenant baseline workload by 30 tenants.

• Goal: Validate system stability and resource isolation under load 30 tenants load. The success criteria focused on maintaining API p95 <2s and ensuring Database CPU remains below 50% while keeping average system resource usage (CPU/Memory) within the optimal 40–60% range.

Phase 3: Hybrid Architecture Validation (Cost Optimization)

The final phase tested an experimental Hybrid Database Deployment to evaluate potential cost reductions without compromising performance.

• Scope: Integration of the provisioned db.r7g.4xlarge instance with Aurora Serverless V2 (1–4 ACU). The workload involved 30 concurrent tenants with intentionally scheduled "burst traffic" (synchronized large loads/DIs) to trigger serverless auto-scaling.

• Goal: Prove that the hybrid model can deliver performance parity to an all-provisioned setup while achieving ≥40% cost savings (compared to a db.r7g.8xlarge). Success relies on the Serverless component scaling within <30 seconds with zero connection errors.

Summary

Test Runs 

Duraion of each test = 30 minutes

The performance validation was executed in three targeted runs:

• Test_1_STB: Single Tenant Baseline
  On a single tenant(cs00000001_0002) and the provisioned db.r7g.4xlarge DB instance.

  Authentication - 100 Users,
  Circulation = ( Check In -50 operation/min, Check Out -50 operation/min),
  Search operation - 200 searches per minute,
  Inventory- 100 requests per minutes)
  
  

• Test_2 _MTC: Multi-Tenant Concurrency (30 Tenants)
  Scaled the baseline workload to 30 concurrent tenants on the segcon environment. This run the same test as on Test_1_STB, but on 30 tenants with 10 different users on each tenant and the provisioned db.r7g.4xlarge DB instance.
  Authentication - 270 Users,
  Circulation = 32300 CICO Transactions per 30 minmutes test. 36 transaction per minute
  Search operation - 237554 searches request per test. 264 searches request per test
  Inventory- 80 requests per minutes
  Data-import 25K, Create profile - 10 imports in parralel on 10 first tenants

• Test_3_HDv1: Hybrid Deployment 4xl +1 xl Executed the 30-tenant workload on the Hybrid architecture (Provisioned db.r7g.4xlarge + Aurora Serverless db.r7g.1xlarge).
  Authentication - 270 Users,
  Circulation = 32400 CICO Transactions per 30 minmutes test. 37 transaction per minute
  Search operation - 241904.00 searches request per test. 264 searches request per test
  Inventory- 80 requests per minutes
  Data-import 25K, Create profile - 10 imports in parralel on 10 first tenants
  
  

Non- functional requirements:

• API p95 <2s, p99 <5s

• Average CPU 40-60%

• Average Memory 40-60%

• Database CPU <50%

• All tenant schemas isolated correctly

Results

Test statuses

Test_1_STB