By Dugsong, via Wikimedia Commons
Some systems such as Infrastructure as a service (IaaS) are similar to traditional software development systems. With these fimilar environmens testers do not have to change much of their methods of testing.
Other systems such as platform as a system (PaaS) can represent entirely new paradigms. Multitendency system and databases, for example, are systems that exist across multiple systems. Traditional test automation can involve inserting and them manipulating data. If the testing system does not pin itself to one database database update latency may cause unexpected errors. PaaS often do not have the same testing hooks that many mature software languages do.
Software as a Service (SaaS) can suffer from high loads. The cloud provider may also not give the developers a test environment to work with. When they do find bugs it may be up to the cloud provider to fix. The cloud provider may not be willing to fix the service before the systems is due to go into production.
Many cloud systems change versions outside of the SDLC of the current project. System that are tested and working may break without notice. Systems with governors may work at the time of release but break under strain.
In short testers are the ones who have the great burden working with could computing. They need to find different ways of testing. In the past testing mainly happened before deployment into production and was expected to remain working the same until the next release. Cloud computing changes that. Systems can easily break after deployment changing the way testers need to test.