Bahmni LITE Infra-cost estimates (based on Performance testing)
- Ramkumar G
- Gurpreet Luthra
- M. Maharaja .
All the estimates projected are based on tests performed for 24 hours on single Kubernetes namespace/single cluster with multiple concurrent users and a shared RDS for Bahmni Lite 1.0 release.
Cost Estimation Sheet with Results
Comprehensive test results and calculation estimates can be viewed in this google sheet document: Bahmni Lite Setup and Cost Estimation
The below analysis explains the various design factors that were taken into consideration during the performance test calculation process.
Smaller cluster (approx 40 concurrent users)
t3.large instance is suggested for Bahmni lite implementation since the number of active pods may vary from “14 to 22” based on the requirement. t3.large supports upto 35 kubernetes pods and enables seamless deployments.
Report
Gatling Stats - Global Information
Patient Records in DB (prior to running the test): 75000
The given hardware supported 40 concurrent users during our test with acceptable performance. The clinical activities performed under the test were
Activities | Total performed | Single user performed |
Patient Created | 5760 | 720 |
Patient Search | 8640 | 720 |
Patient Consultation | 5760 | 288 |
Patient Document Upload | 1440 | 720 |
Based on the above numbers & hardware, we recommend the following:
If the expected patient / traffic load matches the test data then, on the same hardware you can run up to 20 clinics (2 users per clinic) or 10 clinics (4 users per clinic).
If the expected patient / traffic load is LESS than 75% of the above test, then you can likely run up to 24+ clinics (with 2 users-per-clinic) or 12+ clinics (4 users per clinic).
Larger cluster (approx 70 concurrent users)
m5.xlarge instance is suggested for Bahmni lite implementation on this setup since both the CPU and RAM availability is twice than t3.large and supports upto 58 kubernetes pods.
Report
Gatling Stats - Global Information
Patient Records in DB prior to running the tests: 90000
The given hardware supported 70 concurrent users during our test with acceptable performance. The clinical activities performed under the test were
Activities | Total performed | Single user performed |
Patient Created | 10080 | 720 |
Patient Search | 14400 | 720 |
Patient Consultation | 10080 | 288 |
Patient Document Upload | 2160 | 720 |
Based on the above numbers & hardware, we recommend the following:
If the expected patient / traffic load matches the test data then, on the same hardware you can run up to 35 clinics (2 users per clinic) or 17 clinics (4 users per clinic).
If the expected patient / traffic load is LESS than 75% of the above test, then you can likely run up to 40+ clinics (2 users-per-clinic) or 20 clinics (4 users per clinic).
NOTE: The number of users assumed are based on the number of requests sent by a user on an average to the server and typical clinic setup observed from real time sources.
100+ Concurrent Users Long Duration Test Data
For detailed Test results and more scenarios please see this wiki page: Bahmni Lite Performance Long Duration Simulation Baselining
The Bahmni documentation is licensed under Creative Commons Attribution-ShareAlike 4.0 International (CC BY-SA 4.0)