smom-dbis-138/docs/azure/KUBERNETES-36REGION-MAPPING.md

Kubernetes Cluster & Pod Density Mapping - 36-Region Blueprint

Overview

This document maps Kubernetes cluster configurations and pod densities to the 36-region global deployment blueprint.

---

🖥️ VM Configuration per Region

Primary Regions (12 regions) - 8 vCPUs each

Cluster Resources:

• System Node Pool: 2 × Standard_D2s_v3
  • 4 vCPUs total
  • 16 GB RAM total
• Validator Node Pool: 2 × Standard_B2s
  • 4 vCPUs total
  • 8 GB RAM total
• Total Capacity: 8 vCPUs, 24 GB RAM per region

Remaining Regions (24 regions) - 6 vCPUs each

Cluster Resources:

• System Node Pool: 2 × Standard_D2s_v3
  • 4 vCPUs total
  • 16 GB RAM total
• Validator Node Pool: 1 × Standard_B2s
  • 2 vCPUs total
  • 4 GB RAM total
• Total Capacity: 6 vCPUs, 20 GB RAM per region

---

📦 Pod Density Mapping

Pods per Region (All 36 regions)

1. Hyperledger Besu Network

Validators: 4 pods per region

• CPU request: 2 cores per pod
• Memory request: 4 Gi per pod
• Total CPU: 8 cores per region
• Total Memory: 16 Gi per region
• Placement: On validator node pool

Sentries: 3 pods per region

• CPU request: 2 cores per pod
• Memory request: 4 Gi per pod
• Total CPU: 6 cores per region
• Total Memory: 12 Gi per region
• Placement: On system node pool

RPC Nodes: 3 pods per region

• CPU request: 4 cores per pod
• Memory request: 8 Gi per pod
• Total CPU: 12 cores per region
• Total Memory: 24 Gi per region
• Placement: On system node pool

Besu Subtotal: 10 pods per region

2. Application Stack

Blockscout:

• Explorer: 1 pod (1 CPU, 2 Gi mem)
• Database: 1 pod (0.5 CPU, 1 Gi mem)

FireFly:

• Core: 1 pod (1 CPU, 2 Gi mem)
• PostgreSQL: 1 pod (0.5 CPU, 1 Gi mem)
• IPFS: 1 pod (0.5 CPU, 1 Gi mem)

Cacti:

• API Server: 1 pod (1 CPU, 2 Gi mem)
• Besu Connector: 1 pod (0.5 CPU, 1 Gi mem)

Monitoring Stack:

• Prometheus: 1 pod (0.5 CPU, 2 Gi mem)
• Grafana: 1 pod (0.5 CPU, 1 Gi mem)
• Loki: 1 pod (0.5 CPU, 1 Gi mem)
• Alertmanager: 1 pod (0.1 CPU, 0.25 Gi mem)

API Gateway:

• Nginx Gateway: 2 pods (0.1 CPU, 0.128 Gi mem each)

Application Subtotal: 13 pods per region

---

📊 Total Pod Density

Per Region Totals

• Besu Network: 10 pods
• Application Stack: 13 pods
• Total Pods per Region: 23 pods

Global Totals (36 regions)

• Total Pods: 828 pods (36 regions × 23 pods)
• Besu Pods: 360 pods (36 × 10)
• Application Pods: 468 pods (36 × 13)

---

🎯 Pod Placement Strategy

Primary Regions (8 vCPUs capacity)

Validator Node Pool (4 vCPUs, 8 GB RAM):

• Besu Validators: 4 pods
  • CPU request: 8 cores total
  • Memory request: 16 Gi total
  • Note: Requires resource overcommit (4 vCPUs available, 8 cores requested)

System Node Pool (4 vCPUs, 16 GB RAM):

• Besu Sentries: 3 pods (6 cores, 12 Gi)
• Besu RPC: 3 pods (12 cores, 24 Gi)
• Application Stack: 13 pods (~5.2 cores, ~13 Gi)
• Total Requests: ~23.2 cores, ~49 Gi
• Note: Requires resource overcommit (4 vCPUs available, ~23 cores requested)

Remaining Regions (6 vCPUs capacity)

Validator Node Pool (2 vCPUs, 4 GB RAM):

• Besu Validators: 4 pods
  • CPU request: 8 cores total
  • Memory request: 16 Gi total
  • Note: Requires resource overcommit (2 vCPUs available, 8 cores requested)

System Node Pool (4 vCPUs, 16 GB RAM):

• Besu Sentries: 3 pods (6 cores, 12 Gi)
• Besu RPC: 3 pods (12 cores, 24 Gi)
• Application Stack: 13 pods (~5.2 cores, ~13 Gi)
• Total Requests: ~23.2 cores, ~49 Gi
• Note: Requires resource overcommit (4 vCPUs available, ~23 cores requested)

---

⚠️ Resource Overcommit Considerations

Kubernetes Resource Overcommit

CPU Overcommit:

• Validator nodes: 2-4x overcommit (requested vs. available)
• System nodes: 4-6x overcommit (requested vs. available)

Memory Overcommit:

• Validator nodes: 2x overcommit (requested vs. available)
• System nodes: 3x overcommit (requested vs. available)

Best Practices:

• Use resource limits to prevent pod starvation
• Configure QoS classes (Guaranteed, Burstable, BestEffort)
• Monitor actual resource usage and adjust requests/limits
• Consider horizontal pod autoscaling (HPA) for non-critical workloads

---

🔧 Recommended Adjustments

Option 1: Reduce Pod Resource Requests

Reduce Besu RPC CPU requests:

• From 4 cores to 2 cores per pod
• Total: 6 cores per region (instead of 12)
• Helps fit within system node capacity

Reduce Application Stack requests:

• Use smaller resource requests where possible
• Some pods can run with lower CPU allocation

Option 2: Increase System Node Count

Primary regions:

• Increase system nodes to 3 (6 vCPUs, 24 GB RAM)
• Total: 10 vCPUs per region (at quota limit)

Remaining regions:

• Increase system nodes to 3 (6 vCPUs, 24 GB RAM)
• Total: 8 vCPUs per region (within quota)

Option 3: Hybrid Approach

• Use current VM counts (2 system nodes)
• Reduce resource requests for non-critical pods
• Enable HPA for dynamic scaling
• Accept moderate overcommit with proper limits

---

📋 AKS Cluster Configuration

Cluster Settings

cluster_config:
  name: "${region}-aks-main"
  kubernetes_version: "1.32"
  network_plugin: "azure"
  network_policy: "azure"
  
  default_node_pool:
    name: "system"
    node_count: 2
    vm_size: "Standard_D2s_v3"
    os_disk_size_gb: 128
    
  validator_node_pool:
    name: "validators"
    node_count: 1  # or 2 for primary regions
    vm_size: "Standard_B2s"
    os_disk_size_gb: 512
    node_taints:
      - "role=validator:NoSchedule"

---

🚀 Deployment Sequence

Phase 1: Infrastructure Deployment

1. Deploy AKS clusters in all 36 regions
2. Configure system node pools (2 nodes each)
3. Configure validator node pools (1-2 nodes based on region type)
4. Verify cluster connectivity and health

Phase 2: Besu Network Deployment

1. Deploy Besu validators (4 pods per region)
2. Deploy Besu sentries (3 pods per region)
3. Deploy Besu RPC nodes (3 pods per region)
4. Configure geo-aware validator selection
5. Initialize IBFT 2.0 consensus

Phase 3: Application Stack Deployment

1. Deploy Blockscout (explorer + database)
2. Deploy FireFly (core + PostgreSQL + IPFS)
3. Deploy Cacti (API + connector)
4. Deploy monitoring stack
5. Deploy API gateway

Phase 4: Verification & Optimization

1. Verify all pods are running
2. Monitor resource utilization
3. Adjust resource requests/limits as needed
4. Enable autoscaling where appropriate
5. Test geo-aware consensus

---

📊 Resource Summary

Global Totals

Resource	Total (36 regions)
VMs	120
vCPUs	240
RAM	~440 GB
Pods	828
Besu Pods	360
Application Pods	468

Per Region Averages

Resource	Primary (12)	Remaining (24)
VMs	4	3
vCPUs	8	6
RAM	24 GB	20 GB
Pods	23	23

---

📚 References

• 36-Region Blueprint
• Geo-Aware Committee Configuration
• Deployment Checklist