- Selecting the subsystem architecture and topology capable of providing high quality electricity to servers that are co-located is the first major undertake (task 1.1).
- A description of different strategies of monitoring the available power, from different intermittent sources of power will be considered (task 1.2).
- In addition, the robustness of the selected architecture will be verified (task 1.3).
- On the other hand, shifting data from one server to another with both servers under sufficient power availability conditions and subjected to willful electric defaults will be considered (task 1.4).As a result, the system should demonstrate its ability to overcome electric failures and ensure transparent data shifting.
- Then, unpredictable events due to real-world weather constraints will be gradually introduce to the system (task 1.5). For instance, the system has to provide a fast response to an eventual lack of solar power, wind power, and other sources of energy.
- A nonstop run of the system will be monitored in order to detect eventual anomalies (task 1.6). Data from different seasons will be used for a period to be determined.
- Finally, a scaling up design in order to implement a field test will be developed (task 1.7).
Components
From an electric and control stand point, this project will be
divided into seven major tasks(sub modules).
A system emulator will be used to simulate the behavior of real
renewable energy sources. A short description of additional
components is given below:
- a. RES: Renewable Energy Sources: wind turbines and solar panels will be used to respectively transform kinetic motion and light intensity to electricity (respectively 3 phases AC voltage and one DC voltage line);
- b. CC: Charge Controller is the device responsible for rectifying the AC voltage and providing continuous DC voltage. It is also responsible for controlling the output DC intensity current to avoid electric components failure (including servers and other subsystems);
- c. EMU: Energy Management Unit is managing and monitoring energy input/output, state of charge of the storage unit. It is also responsible for managing energy resources in peak times and other critical periods of the day;
- d. SCU: Signal Conditioning Units are a series of power electronic devices such as rectifiers, inverters and UPS. Generally, these components cannot be all used at the same time, however, if the application is critical, they can be combined;
- e. PDU: Power Distribution Unit is responsible for intelligently distributing the power available at the source to the servers based on their computing needs and efficient operations.
- f. Load: The HP ProLiant DL385 G7 servers are currently being selected for this application. It has multi-core processors (4/8/16 µP), each server is dedicated for highly mission critical applications; this server includes 12-core AMD Opteron 6100 Series.
Test Bench Organization
Two electric panels will be used, one with 48VDC and
the other with 208 VAC. Two experiments will be realized as shown in
the below figures.
The below figure below explains a possible way
datacenter can be managed. First, the total data to be process during
a time slot is estimated. Second, the Workload Energy Manager (WEM)
evaluates the necessary amount of electric energy needed to meet the
demand and prepares different scenarios that are sent to a UC solver
which prepares a schedule for each scenario. The WEM receives back
all solutions and stores them in a database. When one of these
scenarios occurs, the WEM orders the switch to move data from one
server to another accordingly.
Testbed Architecture
In our testbed, the actual migration traffic is
isolated from the network we use to control the tests and to monitor
the state of the VM being migrated. A third machine simulates switch
connecting two Hosts on a private network. We can use either an
actual physical switch or a third machine running the CORE network
emulation software to represent a variety of network conditions.