Sizing Considerations for KOBIL Shift deployments
Shift Sizing Considerations
General Information:
ReSizing for KOBIL Shift is always possible at Pods(Replica)- and resource-level (CPU, Memory) update on existing deployment. Any initial deployment could be tuned later without a need for a fresh new deployment by running a KOBIL Shift upgrade. KOBIL expect the Kubernetes Cluster resources could be dynamically reallocated to provide higher resource values (i.e. number of CPUs, amount of RAM, also in case of needs different Node-Hardware-Types). By this any "initial deployment" could be tuned depending on subsequent workload monitoring. The below table is for KOBIL Shift Light. Adding Smart-services and SCP-services increases the resource allocation 30% each.
-
The minimum setup (0 Logins/s) is the smallest runnable configuration (e.g. for test, development etc. environments).
-
IMPORTANT: The login/s is the relevant scaling parameter. For better visibility, the number of users has been added from KOBIL experience with our customer projects.
-
This setup assumes high availability in the sense that there shall be no interruption in backend service if one pod crashes. Therefore 2 pods (replica) per component have been assumed for this setup.
-
HA Implementation for the DBMS is not addressed here (as DBMS is used as a service) and should be calculated seperately.
-
Kafka- and Redis-Services are not covered here as this is part of the platform infrastructure services (overall for this we expect to allocate up to 4 CPUs/Cores and 6GB RAM)
| Number of user (Login/s) | Number of user (Login/s) | Number of user (Login/s) | Number of user (Login/s) | Number of user (Login/s) | Number of user (Login/s) | |||
|---|---|---|---|---|---|---|---|---|
| Application | Resources | 0 (0 Login/s) | 100.000 (2 Login/s) | 1.000.000 (20 Login/s) | 2.000.000 (60 Login/s) | 3.000.000 (60 Login/s) | 5.000.000 (100 Login/s) | 10.000.000 (200 Login/s) |
| KOBIL Shift | Cores total (Worker-Node) | 10 | 10 | 10 | 16 | 22 | 28 | 40 |
| KOBIL Shift | RAM total (Worked-Node) | 12 GB | 16 GB | 32 GB | 48 GB | 52 GB | 64 GB | 100 GB |
| KOBIL Shift | Storage/Diskspace Worker-Node | <40 GB | <40 GB | <40 GB | <40 GB | <40 GB | <40 GB | <40 GB |
| DB PostgreSQL | Cores | 2 | 2 | 4 | 6 | 6 | 8 | 16 |
| DB PostgreSQL | RAM | 6 GB | 6 GB | 10 GB | 10 GB | 10 GB | 10 GB | 20 GB |
| DB PostgreSQL | Diskspace | 10-20 GB per year for records | 10-20 GB per year for records | 10-30 GB per year for records | 10-40 GB per year for records | 10-40 GB per year for records | 10-80 GB per year for records | 10-80+ GB per year for records |