The extensive growth in popularity of Internet of Things (IoT) has led to the generation of massive amount of data from several heterogeneous sensory devices. This has also led to the increase in energy consumption by these connected devices. Smart buildings are one such platform which are equipped with several micro-controllers and sensors, generating a huge amount of redundant information at their data acquisition level. As a result, real-time applications may not be efficiently executed due to latency delays at the cloud service end. This requires several devices at cloud service end to execute the massive amount of data generated by these sensors, which does not satisfy green computing criteria. In this context, a novel local processing mechanism (LPM) is proposed, which favors an improved IoT service architecture for smart buildings. From the perspective of green computing, the proposed LPM framework facilitates reduction of manifolds at data acquisition level of sensor nodes. This paper also addresses the concept of optimal use of sensors in a wireless sensor network (WSN) and estimates costs corresponding to non-Poisson and Poisson arrival of data packets at local processor using the well-known queuing model. We also provide an efficient algorithm for smart buildings using our expert Markov switching (EMS) model, which is a well known probabilistic model in the field of artificial intelligence (AI) for subjectively validating real sensory data sets (viz., temperature, pressure, and humidity). Further, it has been analyzed that the proposed EMS algorithm outperforms several other algorithms conventionally used for determining the state of large-scale dynamic sensor networks. The service cost of proposed model has been compared with conventional model under various stress conditions viz., arrival rate, service rate, and number of clusters. It is observed that the proposed model operates well by leveraging green computing criteria. Thus, in the aforementioned context, this paper provides thing-centric, data-centric, and service-oriented IoT architecture.

物联网（IoT）的广泛普及已导致从几种异构的传感设备中生成大量数据。这也导致这些连接的设备的能耗增加。智能建筑就是这样一种平台，配备了多个微控制器和传感器，在其数据采集级别上会生成大量的冗余信息。结果，由于云服务端的等待时间延迟，实时应用可能无法有效执行。这要求云服务端的多个设备执行由这些传感器生成的大量数据，而这些数据不满足绿色计算标准。在这种情况下，提出了一种新颖的本地处理机制（LPM），该机制支持针对智能建筑的改进的IoT服务架构。从绿色计算的角度来看，提出的LPM框架有助于在传感器节点的数据采集级别上减少流形。本文还讨论了在无线传感器网络（WSN）中最佳使用传感器的概念，并使用众所周知的排队模型估算了与非泊松和泊松到达本地处理器的数据包相对应的成本。我们还使用专家马尔可夫切换（EMS）模型为智能建筑提供了一种有效的算法，该模型是人工智能（AI）领域中众所周知的概率模型，用于主观验证真实的感官数据集（即温度，压力，和湿度）。此外，已经分析出，提出的EMS算法优于常规用于确定大规模动态传感器网络状态的其他几种算法。所提出的模型的服务成本已与常规模型在各种压力条件下进行了比较，即到达率，服务率和集群数量。可以看出，通过利用绿色计算标准，提出的模型运行良好。因此，在上述情况下，本文提供了以事物为中心，以数据为中心和面向服务的物联网架构。