Most organisations I've worked for were scaled N+1 for day-to-day reliability with 2N for disaster recovery. This meant that the total servers required were 2(

Most organisations I've worked for were scaled N+1 for day-to-day reliability with 2N for disaster recovery. This meant that the total servers required were 2(N+1). You get even more idle servers when you consider that they were scaled for New Year's eve traffic, which is typically 10x regular peak traffic (phone system). 30% idle is easily achievable if you're doing proper disaster recovery (recovery when the entire office is gone due to earthquake, flood, fire, etc).