What is the typical "up-time" percentage that should be included in the service level agreement?

Most vendors will provide you this number as part of their service level agreement. Uptime is measured in "9's". For example, an uptime of 99.9% (3-9's) means that they system is will be down about 9 hours per year. An uptime of 99.99% reduces downtime to less than an hour per year. Just as important as uptime is recovery time - the time it takes to recover your data (Recovery Time Objective) and the point in time from which they can recover your data (Recovery Point Objective). For example, if the RTO is 4 and the RPO is 8, you should expect to have your system back up and running in 4 hours BUT your most recent data will be 8 hours old - that means the last time they took a snapshot was 8 hours ago.

Quantifying the costs is a point of contention. Most vendors would like to limit their exposure to the time that you did not have the system - in others words if you are paying $10 per hour to use the system and you lose the system for 1 hour, they would reimburse you for $10.

Of course, your costs can be higher - because you may lose business during the outage. You might be able to withstand an outage of 1 hour but a 1 week outage could put you out of business. The strong-point about cloud computing is that extended options should be a lot rarer than on-premise outages. For example, here on the east coast, many companies were without power for up to two weeks. If they relied solely on on-premise systems, they had a major exposure.

We rely almost exclusively on cloud services for everything from accounting to document management. So while the storm did hamper our business, we lost no data and had access to our systems from wifi hotspots.