Data integrity has to be an important issue for everybody. Sure, some things may be more important than others, but does anybody bother to store data that isn’t of some importance? The matter is important for the reputation of the cloud providers. Even if the things they lost last time were somehow unimportant, wouldn’t we take that as a warning not to trust them next time?
Cloud providers tend to balance cost effectiveness of services with the quality of the services; it’s the same for a lot of business situations. There is a general trend for the client’s level of protection to vary with the type of service. IaaS (Infrastructure as a service) provides a means of creating a cloud environment, but data backup may not be included. PaaS (Platform as a Service) will have more selling points, usually including data protection. But this varies greatly between providers. Some IaaS providers have good solid data protection options, though you will pay a little more for this security.
There is data on the Cloud that is part of the provider’s operations that does not affect the customer to any real degree. Extreme loss here might jeopardise the provider’s business, but really their ability to protect this data is only a concern to us in as much as is indicates their general level of security.
Of more concern is data loss that affects either the provider and the client, or only the client. The first category would include environment matters, Configurations, virtual networking, Provisioning management …etc. This is not the data provided by the client but the packages it is stored in; it’s like losing your word processor program. The Second category is the client’s information, the data they put on the cloud. This is like the words typed into the word processor, but not the program itself. Loss of this is what we are most concerned with.
Really it is the customer who is most responsible for this data. Providers vary in what they offer, so the client has to choose the best options available. There was certainly data loss before the Cloud, and we took measures against it then. We need to be equally active now, or at least use a provider who is active for us. Some measures include:
Disk Level data Protection: An old but effective practice.
Constant backup: Periodically backing up data to a lower cost medium. Somebody has to decide how often to do this, and understand that recent updates will be lost, but otherwise this is a tried and true system. Memory of several Terabytes is quite cheap these days.
Data replication: Another older idea that stays in use because it works well. Software sends all the data to two different storage mediums. But check the ability to retrieve the data from secondary resources.
Journaled/checkpoint based replication.

The cloud infrastructures and provider services are important to look at, but we must remember that the main reason for downtime remains human error. One human error is simply choosing something that isn’t the best suited option.