Proposed Demo Architecture
With different characteristics of Renewable Energy, we can decompose a datacenter into Computation, Network Flow, and Data access. Computing servers consume most of the power but can be leveraged by Virtualization or other Computer Techniques to shift the services across datacenters while network and storage equipment take less power. This is critical to the users, but needs long-term access so computing servers can be supplied with wind turbines or hydro dams which provide the power at KWs level, while network and storage equipment can be supplied with Solar panels which provides the power at 100Ws. For computing servers, as the renewable energy is not dynamic and inconstant, Virtualization can be leveraged in both shortage and outage cases and Service Scheduling Components will be in place to manage the computing services: 1. In power shortage cases, Virtualization can dynamically consolidate the services in fewer Servers while shutting down some servers 2. In power outage cases, Virtualization can dynamically migrate the services to another Datacenter to escape the shutdown datacenter For storage equipment, the low-powered but long-lasting renewable energy can still maintain the data access through network. So the proposed Demo Architecture will look like the diagram as below:
Proposed Demo Architecture
In the above proposed architecture: Based on the proposed architecture, the protocol between Service Scheduler and Power Predicator is focused on how quickly the power predictor can notify the Service Scheduler of the power outage/shortage and how long the power supply can last in current strength. Those two statistics are critical for enabling the Service Scheduler to make a good decision about migrating the computing service.
Proposed Hardware Specifications
With the current project budget, we plan the following: each rack-mount server represents a mini-datacenter with virtualization deployed. One mini-datacenter supplied with Green Energy, and the other mini-datacenter supplied with Grid Power. Power Predictor will predict the power of Green-energy-driven datacenter and Service scheduler will decide when to shift the computing service onto and off GDC. This system will include: