Scaling and Performance

Getting the Most Out of Your Instance

This page covers how to monitor resource usage, recognize when you need more capacity, and optimize performance within your current tier. Whether you are on ClawHosters or self-hosting, these principles apply.

Understanding Resource Limits

Each ClawHosters tier provides specific resource allocations:

The container memory limit is enforced by Docker. If your OpenClaw process exceeds this limit, Docker restarts the container automatically. The host RAM is shared between the operating system, Docker overhead, and the container.

Signs You Need to Upgrade

Memory Pressure

Your instance may need more memory if you observe:

• Frequent container restarts -- Check with docker inspect openclaw | grep RestartCount or look at the dashboard status history
• Slow response times -- When memory is nearly full, the system spends more time managing memory than processing requests
• Out-of-memory kills -- Docker logs will show OOM (out of memory) events

CPU Saturation

Your instance may need more CPU if you observe:

• Consistently slow responses even when memory usage is normal
• Long processing times for tasks that involve file operations or web automation
• Playwright timeouts -- Browser automation is CPU-intensive and may timeout on smaller tiers

Storage Running Low

Check storage usage on self-hosted instances with:

docker system df
df -h

On ClawHosters, storage usage is visible in the dashboard.

Monitoring Your Instance

On ClawHosters

The dashboard provides basic monitoring:

• Instance status (running, stopped, restarting)
• Uptime since last restart
• Current tier and resource allocation

On Self-hosted Instances

Use standard Linux tools to monitor resource usage:

# Memory usage
free -h
docker stats openclaw --no-stream

# CPU usage
top -bn1 | head -20

# Disk usage
df -h
docker system df

Container-Level Monitoring

# Real-time resource usage
docker stats openclaw

# Check restart count
docker inspect openclaw --format='{{.RestartCount}}'

# View recent logs for errors
docker logs openclaw --tail 100

Optimizing Performance

LLM Response Times

The largest factor in response speed is usually the LLM provider, not your instance resources:

To improve LLM response times:

• Use a simpler model for routine tasks -- Not every query needs the most capable model
• Keep conversation context manageable -- Very long conversations slow down each subsequent response
• Start new conversations periodically -- This resets the context window

Memory Optimization

Reduce memory usage without upgrading:

• Limit installed packages -- Each package in the container uses memory
• Reduce Playwright usage -- Browser instances consume significant memory. Close them when not in use
• Monitor conversation history size -- Large conversation logs accumulate over time

Web Automation Performance

Playwright (browser automation) is the most resource-intensive feature:

• Budget tier: Can run basic web tasks, but complex pages may timeout
• Balanced tier: Handles most web automation tasks reliably
• Pro tier: Best for heavy web automation workloads

Tips for better Playwright performance:

• Avoid running multiple browser tabs simultaneously on Budget tier
• Use headless mode (default) for lower resource consumption
• Close browser contexts after completing tasks

Upgrading Your Tier

On ClawHosters

To upgrade your tier:

1. Go to your instance settings in the dashboard
2. Select a higher tier
3. The instance will be rebuilt with the new resources
4. Your data volume is preserved during the upgrade

Note: Upgrading requires a brief restart. Your data and configuration are preserved because they are stored on the persistent data volume.

Downgrading

Downgrading works the same way. Be aware that if your current memory usage exceeds the lower tier's limit, the container may restart more frequently after downgrading.

On Self-hosted Instances

To scale a self-hosted instance:

1. Vertical scaling: Upgrade your VPS to a larger plan (more CPU/RAM)
2. Memory limit: Update the mem_limit in your docker-compose.yml
3. Restart the container to apply the new limits

# docker-compose.yml - increase memory limit
services:
  openclaw:
    mem_limit: 4g  # Increase from 2g to 4g

When Scaling Does Not Help

Some performance issues are not solved by adding resources:

• LLM provider latency -- Response time depends on the API provider, not your server
• Network issues -- Slow webhooks or API calls are network-bound, not CPU-bound
• Plugin bottlenecks -- A poorly written plugin can slow things down regardless of resources
• Rate limits -- LLM providers impose rate limits that are independent of your instance size

For these cases, see Troubleshooting Performance for specific solutions.

Related Docs

• Architecture Overview -- How the infrastructure works
• Resource Limits -- Detailed resource allocations per tier
• Troubleshooting Performance -- Diagnosing slow responses
• Docker Configuration Reference -- Container configuration options