Fines data centre ecosystem

All4Green

All4Green Logo

10% - 20% energy saving

End User / Data Centre Sub-ecosystem Components

Thu, 2014-02-27 17:12 -- MehdiAben

This issue refers to the sub-ecosystem comprising the DC and its end users, in order to provide the description of this sub-ecosystem components. It contains the functional and formal definition GreenSLAs, a description of the behaviour of the agents that are relevant in this context (DC Agent and ITC Agent), the policies that determine the actions to be taken, and the necessary interfaces for communication between the different entities. Besides, a description of the DC’s data model according to the requirements of the DC Agent and the DC Connector –the entity that isolates specific DC characteristics from the DC Agent– is made.
In order for a DC to maximize its economic profit within the energy-aware collaborative context of All4Green, the DC managers need to study the load patterns of their IT customers to come up with adequate proposals of GreenSLAs to be offered to its customers. Only via a thorough assessment phase and by monitoring the load history specifically, can the DC manager extrapolate the contents of the GreenSLA that are most effective. On the other hand, it would also be useful if the DC manager were open to receive proposals for new flexibilities from the IT customers, given that the IT customers have the most precise knowledge of what specific IT services they execute internally and, besides, they can have a more precise idea on the expected application load. In conclusion, there is no single, out of the shelf, solution for every DC: both assessment (by monitoring data) and collaboration (between DC manager and IT customers) are needed to tailor the most adequate GreenSLAs for each specific case.
Concerning the economic reward/penalty game established between the DC and its IT customers, we must not forget that the economic implications of executing flexibilities to achieve a certain power profile must be nonnegative for the DC, taking also into account the corresponding reward/penalty game between EP and DC. This means that, if the modification of IT service conditions implies a reduction of revenues received from the IT customers, then the DC should not obtain a smaller reward from the EP. The reward from the EP could be expressed in several manners, such as a direct monetary discount or a reduction of the energy price for a certain period of time –in the second case, the reward should also take into account the PUE of the DC–. On the other hand, if the DC receives a request from the EP to temporally increase the DC’s power consumption, then the reward from the EP should at least compensate the cost of the additional energy that is consumed if there is no extra revenue from the IT customers –this would depend on the specific GreenSLAs–.
The main driver of long-term energy saving is the modification of the categorizable parameters defined by the GreenSLA during the busy hour. Although shifting workload in time has a modest impact on the long-term energy consumption, it re-distributes power consumption along the time axis, reducing power consumption during the ES intervals.

Request an implementation

Reduce Energy up to 20%