Overview

Glean’s Data Analysis feature uses an auto-scaling architecture for both AWS and GCP customers. The system efficiently manages resources by scaling up to support up to 50 concurrent users and spinning down sandboxes after 30 minutes of inactivity.

Expected monthly costs:

  • Base level: $30-45/month
  • With 10 concurrent users (1 hour daily usage): $45-60/month

Technical Architecture

Infrastructure Components

The system operates using shared core machines for sandbox and orchestrator pods, optimized for:

  • Short bursts of high CPU and memory usage
  • Idle state between analytical queries
  • Efficient resource allocation

Base Infrastructure

In steady state, the system runs:

  • 2 sandboxes
  • 1 orchestrator
  • Total base cost = 3 × (cost per machine) per month

Auto-scaling Behavior

The auto-scaling algorithm maintains sandbox utilization at approximately 75% for optimal performance and cost efficiency.

When usage increases:

  1. System scales up beyond 2 sandboxes
  2. Additional machines are provisioned as needed
  3. Costs increase proportionally to usage
  4. System scales back down when demand decreases

Cost Analysis

Example Scenario

Consider this high-usage scenario:

  • 10 users running concurrent analysis
  • 1 hour daily usage
  • All usage occurring simultaneously (worst-case scenario)
1

Resource Allocation

14 machines running to maintain 75% utilization target

2

Cleanup Process

  • Sandboxes marked for deletion after 15 minutes of inactivity
  • Machines downscaled 15 minutes after marking
3

Extra Resource Usage

  • 12 extra sandboxes (14 total - 2 base)
  • Running for 1.5 hours daily
  • Monthly extra hours: 12 × 1.5 × 30 = 540 hours

Cost Breakdown by Provider

Important Notes

All prices referenced are based on us-east regions. Costs may vary slightly across different regions.

Scaling Limits

For questions about costs or scaling limits, please contact your Glean representative.