diff --git a/q4-scalability/architecture.png b/q4-scalability/architecture.png
new file mode 100644
index 0000000..73b2fa2
Binary files /dev/null and b/q4-scalability/architecture.png differ
diff --git a/q4-scalability/calculations.md b/q4-scalability/calculations.md
new file mode 100644
index 0000000..104a5bf
--- /dev/null
+++ b/q4-scalability/calculations.md
@@ -0,0 +1,54 @@
+# Scalability Calculations
+
+## 1. Required Number of Containers
+
+### Given
+
+- Expected traffic: **15,000 requests/second**
+- Capacity per container: **500 requests/second**
+- Safety margin: **30%**
+
+### Calculation
+
+First, calculate the minimum number of containers required:
+
+```text
+15,000 ÷ 500 = 30 containers
+```
+
+Add a 30% safety margin:
+
+```text
+30 × 1.30 = 39 containers
+```
+
+### Result
+
+**39 application containers** are required to handle the expected traffic while maintaining a 30% capacity buffer for traffic spikes and failover scenarios.
+
+---
+
+## 2. Cold Start Strategy
+
+To avoid initial delay during scaling, the following approach is recommended:
+
+- Maintain a certain number of replicas for the application at any moment.
+- Configure the horizontal auto-scaling feature based on CPU usage or requests per second.
+- Utilize Docker containers with minimal size.
+- Pre-pull container images onto worker nodes before deployment.
+
+In that way, initial delay is avoided while the application is scaled properly during traffic peaks.
+---
+
+## 3. Using Ghaymah Block Storage for Stateful Workloads
+
+Ghaymah Block Storage is used to provide stateful storage for applications like databases.
+
+In this design, the database stores its state on Ghaymah Block Storage instead of storing it on the filesystem inside the container. This way, it makes sure that:
+
+- The data will not be lost even after restarting the container/node.
+- It will still be accessible even after rescheduling containers.
+- There will be decoupling between the stateful workloads and the stateless application pods.
+- The storage will be reliable and durable for application data.
+
+Examples of stateful workloads are relational databases, NoSQL databases, and others.
\ No newline at end of file