Aggregation-Based Colocation Datacenter Energy Management in Wholesale Markets
In this paper, we study how co-location datacenter energy cost can be effectively reduced in the wholesale electricity market via cooperative power procurement. Intuitively, by aggregating workloads and renewables across a group of tenants in a collocation datacenter, the overall power demand uncertainty of the colocation datacenter can be reduced, resulting in less chance of being penalized when participating in the wholesale electricity market. We use cooperative game theory to model the cooperative electricity procurement process of tenants as a cooperative game, and show the cost saving benefits of aggregation. Then, a cost allocation scheme based on the marginal contribution of each tenant to the total expected cost is proposed to distribute the aggregation benefits among the participating tenants. Besides, we propose proportional cost allocation scheme to distribute the aggregation benefits among the participating tenants after realizations of power demand and market prices. Finally, numerical experiments based on real-world traces are conducted to illustrate the benefits of aggregation compared to non cooperative power procurement.
This work is to reduce the growing electricity bills of datacenters, from the demand side, substantial efforts have been made, ranging from hardware such as energy-efficient servers, storage devices, and network switches, to software such as virtualization and dynamic CPU speed scaling and capacity provisioning, which have led to dramatic improvements in the energy-efficiency of datacenters. On the other hand, it is also important for datacenters to manage their energy cost from the supply side. As large consumers, datacenters typically have multiple options to procure electricity to meet their power demand. For instance, a datacenter may purchase power from a retailer such as a local utility company with a pre-specified rate by signing bilateral contracts beforehand, or operate by leveraging on-site power generators and energy storage systems.
• Datacenters vulnerable to high deviation penalties due to their highly uncertain power demand.
• It is imperative for datacenters to mitigate risks associated with these sources of uncertainty in order to maximize the cost saving in procuring power from the wholesale electricity market directly.
To incentivize aggregation and distribute aggregation benefits among tenants, this work propose to use cooperative game theory. Specifically, the problem can be formulated into a cooperative game with transferrable cost. In this game, the set of players is the set of tenants who seek to cooperate in reducing electricity cost. We first prove that coalitional formation can reduce energy cost compared to individual power procurement in the wholesale electricity market. Then our cooperative game is shown to be balanced and therefore has a nonempty core. Given that the two existing cost allocation methods, the Shapley value and nucleolus, are not applicable to our game, we design an efficient cost allocation scheme that can guarantee mutual benefits for all participating tenants such that no one has the incentive to break up from the coalition and thus locate a cost allocation in the core. Besides, we discuss how to allocate the cost to each tenant after realizations of power demand and market prices. As the cost function of our cooperative game is defined in expectation, there might be some days such that the participating tenants need to pay more compared to the realized cost. Therefore, coalitional members may choose to deviate from such coalition if overpayment keeps occurring. Therefore, we propose a cost allocation method based on the proportion of the realized cost on every day to ensure that in the long run, the allocated realized cost on average will approach the expected cost almost surely.
In this paper, we have proposed a new approach to minimize the electricity cost for tenants in co-location datacenters participating in the wholesale electricity market. The electricity cost can be effectively reduced by bidding in the day-ahead market collectively since aggregation can reduce the uncertainty of net power demand. We model this aggregation process as a cooperative game and present a cost allocation mechanism based on the marginal contribution of each tenant to the total expected cost to distribute the optimal expected cost to each tenant within the grand coalition. Moveover, we have discussed how to share the coalitional cost after the realizations of net power demand and market prices. Our proposed proportional cost allocation method can ensure the stability of our cooperative bidding game after realizations in the long run. Finally, simulations based on real-world traces verify the effectiveness of our proposed cost saving method.