AVD Scaling: Time-Based vs Demand-Based
The Smart Scaler Team
There are two fundamental approaches to AVD scaling: time-based and demand-based. Each has strengths and weaknesses.
Time-Based Scaling
Scale capacity according to a schedule.
Example:
- 7am: Start 50 hosts
- 9am: Ensure 80 hosts running
- 6pm: Scale down to 20 hosts
- 10pm: Scale down to 2 hosts
Pros
- Predictable
- Simple to understand
- Capacity ready before users arrive
- Works without usage data
Cons
- Doesn’t adapt to unusual days
- May over-provision on quiet days
- May under-provision on busy days
- Requires manual adjustment for anomalies
Demand-Based Scaling
Scale capacity based on actual usage metrics.
Example:
- When average CPU > 70%: Add hosts
- When available sessions < 10: Add hosts
- When average CPU < 30%: Remove hosts
- When utilisation < 50%: Remove hosts
Pros
- Adapts to actual demand
- Handles unexpected spikes
- Handles unexpected quiet periods
- No manual adjustment needed
Cons
- Reactive, not proactive
- Users may wait during scale-up
- Can be “twitchy” without proper thresholds
- Requires good monitoring data
The Best Approach: Combine Them
Smart scaling uses both:
- Time-based baseline: Ensure minimum capacity during expected busy periods
- Demand-based adjustment: Scale up/down from baseline based on actual usage
This gives you:
- Capacity ready when users arrive (time-based)
- Adaptation to actual demand (demand-based)
- Protection against unexpected spikes (demand-based)
- Efficient scale-down when quiet (demand-based)
Example Combined Schedule
Weekday:
- 7am: Minimum 30 hosts (time-based)
- 7am-6pm: Scale between 30-100 based on demand (demand-based)
- 6pm: Minimum drops to 5 hosts (time-based)
- 6pm-7am: Scale between 0-20 based on demand (demand-based)
Weekend:
- All day: Minimum 0 hosts, scale 0-10 based on demand
What The Smart Scaler Does
We use combined scaling with additional intelligence:
- Schedule windows set your time-based baseline
- Host selectors let you target specific hosts
- Holiday awareness adjusts baselines automatically
- Start VM on Connect handles edge cases
The result: optimal capacity at all times, minimal waste.
Want intelligent scaling without the complexity? Try The Smart Scaler free.