User subscription-based resource management for Desktop-as-a-Service platforms

摘要

The Desktop-as-a-Service (DaaS) idiom consists of utilizing a cloud or other server infrastructure to host the user's desktop environment as a virtual desktop. Typical for cloud and DaaS services is the pay-as-you-go pricing model in combination with the availability of multiple subscription types to accommodate the needs of the users. However, optimal cost-efficient allocation of the virtual desktops to the infrastructure proves to be a combinatorial NP-hard problem, for which a heuristic is presented in the current article. We present a cost model for the DaaS service, from which a revenue of different configurations of virtual desktops to the servers can be derived. In this cost model, both subscription fee and penalties for degraded service are recorded, that are described in service-level agreements (SLAs) between the service provider and the users, and make realistic assumptions that different subscription types result in particular SLA contracts. The heuristic proposed states that for a given user base for which the virtual desktops (VDs) must be hosted, the VDs should be spread evenly over the infrastructure. Experiments through discrete event simulation show that this heuristic yields an approximation within 1 % of the theoretically achievable revenue.