HPL ELECTRIC & POWER

Limited | Gurugram, Haryana

Infrastructure Requirements

Phase 1: MDM Platform | Rev 4 | Appendix D

← Back to Proposal

Infrastructure Sizing for 100,000 Meters

This document provides detailed infrastructure specifications to support 100,000 smart meters with full MDM functionality. The architecture supports multi-tenant deployment (shared infrastructure across DISCOMs) and on-premise deployment (mandatory for utilities that do not accept cloud).

REV 4 UPDATES Multi-Tenant Architecture for shared DISCOM deployment | On-Premise Specs (mandatory option) | GCP Delhi Region added | Cost Sharing Model for multi-DISCOM

01 Cloud Deployment (Recommended)

Cloud deployment is recommended for optimal scalability, reduced upfront costs, and automatic failover. All major Indian cloud regions are supported.

100K METERS Production Environment Specification

Component Specification Quantity Purpose
APPLICATION TIER
API Servers 4 vCPU, 16 GB RAM, 100 GB SSD 3 instances REST APIs, Business Logic
Web Servers 2 vCPU, 8 GB RAM, 50 GB SSD 2 instances Dashboard, Consumer Portal
VEE Engine 8 vCPU, 32 GB RAM, 100 GB SSD 2 instances Validation, Estimation, Editing
Worker Nodes 4 vCPU, 16 GB RAM, 50 GB SSD 3 instances Background Jobs, Reports
DATA TIER
PostgreSQL + TimescaleDB 8 vCPU, 64 GB RAM, 2 TB SSD (IOPS: 10,000+) Primary + Replica Time-series data, Master data
Redis Cache 4 vCPU, 16 GB RAM 2 instances (cluster) Caching, Session, Queues
Apache Kafka 4 vCPU, 16 GB RAM, 500 GB SSD 3 brokers Event streaming, HES data ingestion
SUPPORTING SERVICES
Load Balancer Application Load Balancer (managed) 1 Traffic distribution, SSL termination
Object Storage S3/Blob Storage 500 GB initial Reports, Backups, Logs
CDN CloudFront/Azure CDN 1 Static assets, Mobile app APIs

Estimated Monthly Cloud Costs (India Region)

☁
AWS (Mumbai)
₹1.8-2.2L/month

EC2 + RDS + ElastiCache + MSK

☁
Azure (Central India)
₹1.7-2.1L/month

VMs + PostgreSQL + Redis + Event Hubs

☁
GCP (India) RECOMMENDED
₹1.6-2.0L/month

GCE + Cloud SQL + Memorystore + Pub/Sub

Mumbai (asia-south1) | Delhi (asia-south2)

✓ Auto-Scaling Included

The architecture includes Kubernetes-based auto-scaling. During peak hours (billing cycles, mass reads), additional instances spin up automatically. During off-peak, they scale down - optimizing costs. You only pay for what you use.

💡 Cost Estimate Assumptions

Above estimates are based on 1-year reserved instance pricing and optimized resource allocation. On-demand pricing would be approximately 40-50% higher. We recommend committing to reserved instances from deployment for maximum cost efficiency.

02 Multi-Tenant Architecture (Shared DISCOM Deployment)

HPL operates via trade marketing model - winning DISCOM tenders to supply smart meters + software stack. The MDM platform supports shared cloud infrastructure across multiple DISCOMs with complete data isolation per tenant.

🔗 How Multi-Tenancy Works

🗂
Shared Compute

API servers, VEE engine, worker nodes shared across DISCOMs

🔒
Isolated Data

Separate database schema per DISCOM, no cross-tenant access

💰
Cost Sharing

Infrastructure costs distributed across participating DISCOMs

Multi-DISCOM Cost Sharing Model

Deployment Model Total Infra Cost/Month Cost per DISCOM Savings vs Single
Single DISCOM (100K meters) ₹1.8-2.2L ₹1.8-2.2L -
3 DISCOMs Shared (300K meters total) ₹3.5-4.5L ₹1.2-1.5L ~35% savings
5 DISCOMs Shared (500K meters total) ₹6.0-8.0L ₹1.2-1.6L ~40% savings
10 DISCOMs Shared (1M meters total) ₹10-15L ₹1.0-1.5L ~45% savings

✓ Tenant Isolation Features

  • Database Isolation: Separate PostgreSQL schema per DISCOM with row-level security
  • User Management: Independent user base, roles, and permissions per tenant
  • Branding: Custom logos, colors, and domain per DISCOM portal
  • Configuration: Independent VEE rules, report templates, alert thresholds
  • Billing System: Dedicated integration adapters per DISCOM billing system

03 On-Premise Deployment MANDATORY OPTION

⚠ Many Utilities Do Not Accept Cloud

Several DISCOMs have strict data sovereignty requirements and mandate on-premise deployment. The MDM platform fully supports on-premise installation in DISCOM data centers.

100K METERS On-Premise Hardware Specification

Component Specification Quantity Est. Cost
APPLICATION SERVERS
Application Server Dell/HP: 2x Xeon Gold 6248, 128 GB RAM, 2x 480 GB SSD (RAID 1) 3 servers ₹4.5L each
VEE Processing Server Dell/HP: 2x Xeon Gold 6248, 256 GB RAM, 2x 960 GB SSD (RAID 1) 2 servers ₹6L each
DATABASE & STORAGE
Database Server (Primary) Dell/HP: 2x Xeon Platinum 8280, 512 GB RAM, 8x 1.92 TB NVMe SSD (RAID 10) 1 server ₹18L
Database Server (Replica) Dell/HP: 2x Xeon Gold 6248, 256 GB RAM, 8x 1.92 TB NVMe SSD (RAID 10) 1 server ₹12L
SAN Storage Dell EMC Unity / NetApp: 20 TB usable, 10K IOPS 1 unit ₹15L
NETWORKING & SECURITY
Load Balancer F5 BIG-IP / HAProxy on dedicated hardware 2 (HA pair) ₹8L
Firewall Palo Alto / Fortinet 2 (HA pair) ₹6L
Network Switches Cisco/Juniper 10G ToR switches 2 ₹3L

Estimated Hardware CAPEX (100K meters)

₹75-85 Lakhs

Annual Maintenance (AMC ~18%)

₹12-15 Lakhs/year

Power & Cooling

~15 kVA

✓ On-Premise Deployment Includes

  • Installation Support: Trinesis team on-site for setup and configuration
  • OS & Database Setup: Linux (RHEL/Ubuntu), PostgreSQL + TimescaleDB installation
  • Security Hardening: Firewall rules, SSL certificates, access controls
  • Monitoring Setup: Prometheus + Grafana for infrastructure monitoring
  • Training: DISCOM IT team training on operations and maintenance

04 Scaling Guidelines

The MDM platform is designed for horizontal scaling. Here's how requirements change with meter count:

Meter Count Database Storage API Instances VEE Instances Est. Cloud Cost/Month
50,000 1 TB 2 1 ₹1.0-1.3L
100,000 THIS PROPOSAL 2 TB 3 2 ₹1.8-2.2L
250,000 5 TB 5 3 ₹3.5-4.5L
500,000 10 TB 8 5 ₹6.0-8.0L
1,000,000+ 20 TB+ 12+ 8+ ₹10-15L

📈 Data Volume Calculations (100K meters)

  • Interval Data: 100,000 meters x 96 readings/day x 365 days = ~3.5 billion records/year
  • Storage per year: ~400 GB (compressed) for interval data
  • Events/Alarms: ~50 GB/year
  • 5-year retention: ~2.5 TB total

05 Backup & Disaster Recovery

💾 Backup Strategy

  • Database: Continuous WAL archiving + Daily full backup
  • Retention: 30 days point-in-time recovery
  • Off-site: Cross-region replication (cloud) or tape backup (on-prem)
  • RPO: <15 minutes

🔄 Disaster Recovery

  • Cloud: Multi-AZ deployment, auto-failover
  • On-Prem: Standby DB with streaming replication
  • RTO: <4 hours
  • DR Drill: Quarterly testing recommended

🔒 Confidentiality Notice

This document is confidential and proprietary to Trinesis Technologies Pvt. Ltd. It is shared exclusively with HPL Electric & Power Limited for evaluation purposes only.

Ref: TRIN/PROP/HPL/2026-03-D  |  Classification: Confidential  |  Valid Until: April 30, 2026

🔒

Trinesis Technologies

Confidential Proposal