Servers in datacenter are underutilized due to over-provisioning, which contributes heavily toward high-power consumption of the datacenters. Server consolidation can drastically reduce power consumption of datacenters by reducing the number of active servers. However, server consolidation alone can adversely affect network performance. Recently, wireless datacenters using millimeter-wave communications have been proposed and shown to drastically reduce power consumption of networking equipment. Therefore, server-consolidation strategies should leverage novel datacenter network-architectures to minimize power consumption with minimal impact on performance. We present network-aware bandwidth-constrained server consolidation algorithm called Network-Aware Server Consolidation (NASCon), for wireless datacenters that can reduce the power consumption up to 37%. In wireless datacenter, NASCon server consolidation consumes comparable power compared to consolidation optimized with exhaustive search, while being much less computationally expensive and also improving the network performance. However, due to arrival of new-tasks and completion of existing-tasks, consolidated utilization profile of servers change, which may have adverse effect on overall power consumption over time. To overcome this, NASCon algorithm needs to be executed periodically. We propose a mechanism to find optimal inter-consolidation time and derived a mathematical model for the same. The model reduces the computing overheads for determining the optimal inter-consolidation time enabling real-time implementation.

数据中心中的服务器由于配置过多而未得到充分利用，这在很大程度上导致了数据中心的高功耗。服务器整合可以通过减少活动服务器的数量来大幅降低数据中心的功耗。但是，仅服务器整合可能会对网络性能产生不利影响。近来，已经提出了使用毫米波通信的无线数据中心，并且显示出可以大大降低网络设备的功耗。因此，服务器整合策略应利用新颖的数据中心网络体系结构来最大程度地降低功耗，并且对性能的影响最小。我们提出了一种网络感知带宽受限的服务器整合算法，称为网络感知服务器整合（NASCon），可将功耗降低多达37％的无线数据中心。与通过穷举搜索进行优化的整合相比，在无线数据中心中，NASCon服务器整合消耗的功率相当，同时计算成本更低，并且还改善了网络性能。但是，由于新任务的到来和现有任务的完成，服务器的合并利用率配置文件发生了变化，这可能会对一段时间内的总体功耗产生不利影响。为了克服这个问题，NASCon算法需要定期执行。我们提出了一种寻找最佳内部固结时间的机制，并为此推导了数学模型。该模型减少了用于确定实现实时实施的最佳整合时间的计算开销。