Redshift SLA is determined based on uptime, that is, the time the cluster was in the "Available" state.
For example, if the cluster was running for 14 hours in a day and experienced 5 minutes of downtime, the downtime SLA (%) is calculated as 5 minutes / 840 minutes = approximately 0.59%.
Therefore, if Redshift is not running 24 hours but only during specific periods, the calculation should use the actual uptime as the denominator.
Reference : https://aws.amazon.com/redshift/sla/