Virtual Machine placement, that is assigning Virtual Machines (VMs) to Physical Machines (PMs) in a data centre is an important field of research as it is a difficult problem (NP-Hard) while better placements can generate large savings for companies. In particular, we have shown in a previous work that decision makers can take advantage of a multi-objective presentation to make better reassignments that correspond to their preferences. 
In the context of the Enterprise, the problem is even more complex as the Entreprise is not monolithic but is a composition of several, somewhat autonomous, entities (we call them VCs for Virtual data Centres) through acquisitions and reorganisations. VM placement here needs to consider the preferences or objectives of the VCs in terms of usage of their computing resources. For instance, some groups may want to keep some free resources in their own data centres (e.g., a testing group having the objective to patch quickly any defects found in production); while another group may want its VMs to have priority on a specific resource, wherever its VMs are hosted (e.g., a R&D department running CPU intensive processes). This is to some extent a reverse version of the broking problem in multi-clouds with VCs and capital allocators instead of providers and brokers, but with two main differences: (i) it is a mix of cooperation and competition. (ii) exact costs of placements are not known in advance but only after VCs try to place the VMs that they have been given. 

We present here a preliminary study of the multi-objective VM reassignment problem for the Entreprise. We want to give decision makers a large panel of good and accurate solutions covering the different objectives that make sense for them: in this study we consider electricity consumption, VM migration and reliability costs; while letting each entity evaluate and modify possible reassignments according to its own preferences. We propose E-GeNePi, an adaptation of GeNePi to the Enterprise context. The preliminary results show that E-GeNePi finds in average +114% solutions and gets the best hypervolume for 11 experiments out of 12.