CloudNexus Technology | Cortex Engine Architecture

Ingestion

Streaming CUR files & Cost APIs from AWS, Azure, GCP.

Normalization

Unified Schema mapping. Standardize resource tags & metadata.

Cortex AI

Anomaly detection models, Forecasting, RI/SP Optimization logic.

Action

Dashboards, Slack Alerts, Jira Tickets, Auto-remediation hooks.

How We Normalize Multi-Cloud Data

Each cloud provider has its own taxonomy. AWS has "Instances", Azure has "Virtual Machines". AWS bills by the second, GCP by the minute. The Cortex Engine creates a Unified Resource Model (URM) that abstracts these differences.

// Raw AWS Input

{ "lineItem/UsageType": "USW2-BoxUsage:t2.micro", "product/servicecode": "AmazonEC2" }

// Cortex Normalized Output

{
  "provider": "AWS",
  "category": "Compute",
  "resource_type": "Virtual Machine",
  "spec": "t2.micro",
  "region": "us-west-2"
}

Security & Privacy Architecture

We operate on a Zero-Trust principle. CloudNexus does not require root access or write permissions to your production infrastructure for visibility features.

Read-Only IAM Roles We use cross-account IAM roles with explicit ViewOnlyAccess policies.
Ephemeral Processing Sensitive payload data is processed in memory and never persisted to disk unencrypted.
Private Link Support Enterprise plans support AWS PrivateLink and Azure ExpressRoute connections.

Machine Learning Pipeline

Our specialized models are trained on over $5B of cloud spend data.

Forecasting (ARIMA/Prophet)

We use an ensemble of ARIMA and Facebook Prophet models to predict future spend based on seasonality and growth trends, achieving 96% accuracy on 30-day forecasts.

Anomaly Detection (RCF)

Our Random Cut Forest (RCF) implementation detects statistical outliers in real-time, filtering out expected spikes (like scheduled jobs) to reduce alert fatigue.

Contextual RCA (LLMs)

We leverage Llama-3 class LLMs to analyze CloudTrail and Audit logs alongside billing data, providing human-readable Root Cause Analysis for every cost spike.

System Status All Systems Operational

API Latency (p95) 42ms

Ingestion Throughput 1.2GB/sec

Uptime (L30D) 99.994%

Engineered for Reliability

Financial data demands absolute accuracy and availability. Our infrastructure is built on a multi-region Kubernetes architecture with automated failover and self-healing capabilities.

Multi-AZ Redundancy Data is replicated synchronously across 3 Availability Zones to withstand data center failures.
Polyglot Persistence We use TimescaleDB for time-series metrics, Postgres for transactional data, and Redis for real-time caching.
Chaos Engineering We randomly terminate instances in our staging environment daily to ensure our recovery scripts are battle-tested.

API-First Design

Build your own custom reports or trigger internal workflows.

Webhooks & Events

Subscribe to cost events like budget_exceeded or anomaly_detected to trigger functions in your own infrastructure.

POST /webhooks/receive HTTP/1.1
Content-Type: application/json

{
  "event": "cost_spike",
  "data": {
    "service": "AmazonRDS",
    "amount": 450.20,
    "threshold": 100.00,
    "timestamp": "2024-03-15T10:30:00Z"
  }
}

GraphQL API

Query your billing data with flexibility. Retrieve exactly the fields you need, nested by account, region, or custom tags.

query {
  account(id: "prod-aws") {
    spend(range: "last_7_days") {
      total
      byService {
        name
        cost
      }
    }
  }
}

Built for Scale

Whether you have 10 instances or 100,000, Cortex handles the load.

The Cortex Engine Architecture