Data Center Efficiency & Scalability

Nimbula’s Director of Product Management will discuss Nimbula’s learnings over the last few years on how to build and operate a scalable and efficient IaaS cloud. The talk will explain how to design the datacenter, business operations, and applications around the cloud model of IT so that you can achieve an extreme improvement in IT’s scalability while also significantly lowering operational expenses.

Principles of scale
* Building the cloud hierarchically to avoid bottlenecks
* Simplifying cloud networks by flattening them and having a more flexible network isolation model
* Decoupling unrelated concepts such as broadcast domains from tenant isolation
* Scaling out services horizontally to eliminate bottlenecks and single points of failure

Automation for operational efficiency
* Zero-touch installation of cloud nodes
* Self-healing control plane that ensures a constantly available API without heavy administrator interaction
* Automation of placement of newly provisioned resources based on infrastructure tagging, end user demands, and administrative policy

Organizational efficiency
* Eliminating red tape through self-service of compute, storage, and networks – making sure that tenants can be their own administrators from business logic to network topology, to application management
* Multi-layer delegation – distribute authority and drive decisions on access to the lowest level possible so that decisions made by the people with the actual responsibility for the data, the applications, and the business results
* Cross-tenant collaboration – Enable cloud tenants to work together securely, quickly, and without the need for arbitration by the cloud administrator
* Decoupling roles – Make sure that any role – cloud administrator, tenant administrator, and end user can do their work without the need to wait on work from another role to proceed. Have end users work within predefined constraints but have complete freedom within those boundaries.

Application architecture
* Stateless scale-out applications – Not all applications take advantage of the cloud. By eliminating state from the bulk of the application instances and by scaling out applications, you can decouple the operation of the application from the internals of the infrastructure, freeing the application administrator to manage the applications without assistance from the cloud administrator