Any Rolling Stones fans out there? Well I guess if you were singing along to this when it came out, then you didn’t know that you’d be a least privilege geek in 2012 either. Either way, as I was humming along to myself the other day I couldn’t help but think of the metaphor as it relates today to cloud computing’s greatest challenge: secure multi-tenancy.
The ultimate question of security in the cloud revolves around whether or not multiple tenants can coexist without one having any access to the other’s data or applications. Truly guaranteed, secure multi-tenancy has been labeled as both somewhat unattainable and the basis by which every cloud vendor’s security should be measured. What’s interesting about the secure multi-tenancy discussion is that it isn’t exclusive to separate companies sharing the same public cloud infrastructure. It turns out that this is as big an issue for private cloud implementations where cross division or department privacy is required either for governance or compliance reason.
Why is secure multi-tenancy in the cloud the elusive unicorn? The short answer centers on the observation that the cloud’s greatest strength is also potentially its greatest weakness. Sharing under-utilized resources and paying on a metered or “as-used” basis is a fantastic way to leverage existing investments, control costs, and handle the natural ebbs and flows in capacity planning that usually plagues IT. The challenge comes in when two different organizations with different compliance and security policies are sharing the same resource. How does the cloud provider, even internally for private clouds, ensure that nothing spills from one virtual environment into the other on the physical intersection point, the server?
The sheen of sophistication, the wow factor of something new, dazzles our senses somewhat, and subsequently we invest way too much faith in something—we not only put the cart before the horse, we turn it into a hand cart that we think we can push ourselves. Again we see how human nature is the weak link.