OVDC—Omniverse Derived Cache

OVDC is a persistent, distributed key-value store used as a non-volatile cache for the Omniverse real-time rendering pipeline, UJITSO (Universal Just-In-Time Scene Optimization). It durably stores derived data—GPU-ready artifacts computed from source OpenUSD assets—so it does not need to be recomputed. It trades network bandwidth for compute time by caching pre-generated data that multiple GPUs can share.

OVDC was formerly named DDCS (Derived Data Cache Service). The rename to OVDC landed in release 6.0.0 and carried no functional change.

Overview

Derived data generation is computationally expensive. During initial scene loads, GPUs generate derived content as they encounter assets for the first time. OVDC stores this generated data, making it immediately available to other GPUs. During warm scene loads, render workers read all derived content from OVDC instead of regenerating it, significantly reducing load times.

When to Use OVDC

• Running distributed simulation and rendering workloads

• Need to cache derived content across multiple render workers

• Want to reduce scene load times for warm scene loads

Documentation

• Configuration Guide - Helm values, scaling, storage, and monitoring setup

• Deployment Guide - Installing OVDC using Helm

• TLS Configuration - Enable encrypted communication

• Runbooks - Troubleshooting common operational issues

• OVDC: Configure
  • Base Configuration
  • Base Configuration Field Reference
  • Complete Configuration Reference
  • 1. Provision and Scale
    • AWS EKS Example
  • 2. Memory Configuration
  • 3. Volume Configuration
  • Example Kubernetes StorageClass for AWS
  • 4. Telemetry
  • Configuration Recommendations
  • Summary
• OVDC: Deployment
  • Configuration
  • Deployment Steps
  • 1. Configure Helm
  • 2. Prepare Namespace
  • 3. Create Required Secrets
  • 4. Install Helm Chart
  • Uninstall Instructions
  • Summary
• OVDC: Helm Values
  • Top-Level Field Reference
• OVDC: TLS Configuration
  • 1. Create the TLS Secret
  • Self-Signed Certificate (Reference Only)
  • 2. Enable TLS in values.yaml
    • settings.grpcTls Field Reference
  • 3. Optional: Configure CA Root Certificate
  • 4. Complete Example Configuration
  • 5. Apply Configuration Changes
  • 6. Verify TLS Configuration
  • Summary
• OVDC: Metrics Reference
  • Cache and Storage Metrics
  • Error Metrics
  • RocksDB Property Gauges
  • API and gRPC Metrics
  • Kubernetes Metrics Referenced by the Runbooks
  • RocksDB Terms Used in the Runbooks

Runbooks

• OVDC Runbooks
  • Available Runbooks
• OVDC: Cache Misses and Performance Degradation
  • Overview
  • Symptoms and Detection Signals
    • Visible Symptoms
    • Log Messages
      • Write Buffer Pressure
    • Metric Signals
      • Read Path Metrics (Cache Hit/Miss)
      • Error Metrics
      • I/O Performance Metrics
      • Write Path Metrics
  • Root Cause Analysis
    • Possible Causes
      • Cold Cache (Scene or Assets Not Preloaded)
      • Cache Eviction Due to Memory/Disk Pressure
      • Storage Medium has Insufficient Performance
    • Other Possible Causes
  • Troubleshooting Steps
    • Diagnostic Steps
      • 1. Check Cache Hit/Miss Ratios
      • 2. Check Cache Hit/Miss and Disk Seek Metrics
      • 3. Check RocksDB’s Own Stall and Compaction Stats
      • 4. Check I/O Performance Metrics
      • 5. Review Storage Class and Volume Configuration
    • Other Diagnostic Actions
  • Prevention
    • Proactive Monitoring
    • Capacity Planning
• OVDC: No Active Shards
  • Overview
  • Symptoms and Detection Signals
    • Visible Symptoms
    • Log Messages
      • 0 Active Shards Log Message
      • DNS Discovery Failed
    • Metric Signals
  • Root Cause Analysis
    • Known Causes
      • Replica Set Scaled to 0
      • Not Authorized for Nvidia Image repository
      • PVC Attachment Failures
      • Misc Scheduling Issue
    • Other Possible Causes
  • Troubleshooting Steps
    • Diagnostic Steps for Known Root Causes
      • 1. Check if OVDC StatefulSet is Scaled Down to 0
      • 2. Check for Image Pull Authorization Issues
      • 3. Check for PVC Attachment Failures
      • 4. Check for Resource Scheduling Issues
    • Other Diagnostic Actions
  • Prevention
    • Proactive Monitoring
• OVDC: Disk Space Exhaustion
  • Overview
  • Symptoms and Detection Signals
    • Visible Symptoms
    • Metric Signals
      • Volume Capacity Metrics
      • RocksDB Write Stalls from Disk Pressure
      • Database and I/O Performance Metrics
      • Cache Performance Impact
  • Root Cause Analysis
    • Known Causes
      • Insufficient Persistent Volume Allocation
      • Data Growth Outpacing Cleanup
    • Other Possible Causes
  • Troubleshooting Steps
    • Diagnostic Steps for Known Root Causes
      • 1. Monitor Disk Usage Metrics and Set Up Alerts
      • 2. Expand Persistent Volume Claims (PVCs) as Needed
    • Other Diagnostic Actions
  • Prevention
    • Proactive Monitoring
    • Configuration Best Practices
    • Capacity Planning
• OVDC: Network Bandwidth or Latency Bottlenecks
  • Overview
  • Symptoms and Detection Signals
    • Visible Symptoms
    • Metric Signals
  • Root Cause Analysis
    • Known Causes
      • Insufficient OVDC Pods Scheduled
      • Insufficient Network Bandwidth
      • Network Congestion
    • Other Possible Causes
  • Troubleshooting Steps
    • Diagnostic Steps for Known Root Causes
      • 1. Check OVDC Pod Count and Scaling
      • 2. Monitor Network Metrics
    • Other Diagnostic Actions
  • Prevention
    • Proactive Monitoring
• OVDC: RocksDB Corruption or Failures
  • Overview
  • Symptoms and Detection Signals
    • Visible Symptoms
    • Log Messages
      • Database Open Failures
    • Metric Signals
  • Root Cause Analysis
    • Known Causes
      • Unclean Shutdowns
      • Disk Errors or Hardware Failures
    • Other Possible Causes
  • Troubleshooting Steps
    • Diagnostic Steps for Known Root Causes
      • 1. Delete All PVCs and Pods to Force Restart
  • Prevention
    • Proactive Monitoring