At some point in our lives, we've probably all dealt with a noisy roommate or neighbor. All you want to do is go to sleep, but they decided to hold an obscenely loud party that ends up running til 3 a.m. The next day, you wake up tired, groggy and angry.
It's the same way with some cloud providers.
In an unstructured cloud environment, if activity by one or more tenants spikes significantly, it can cause a drain on the overall performance of the hosted solution – which creates that same frustration.
Because of another tenant using up too many iOps or network and storage resources, other companies see their apps and other solutions run slower than they should.
The key to avoiding this problem is guaranteeing clear cloud partitions – with enough dedicated resource space between them that noisy neighbors can't slow your performance. That buffer is like an acre of trees between you and your neighbor down the hall, so even peak workloads aren't an issue.
The second step is to make sure cloud providers aren't just blowing smoke – they should be able to back that up with 100% service level agreements and performance reporting.
With that in place, companies can feel better about migrating mission-critical applications, and I mean Tier 0 or 1 kinds of workloads, to the cloud and avoid the hesitancy they may initially feel about ensuring the performance they need.
It's too bad that kind of segmentation can't work in real life. If it did, it'd be like putting your roommate or "that guy talking way too loudly on his cell phone" into a soundproof booth.