With the current emphasis on intelligent infrastructures, sensor based ubiquitous intelligent systems, commonly known as ‘Cyber-Physical Systems’, have become important. Data acquisition, management, and analysis for knowledge extraction will give rise to a new generation of infrastructure and services encompassing every aspects of our daily lives. Such analysis are performed by well-known algorithms, having various constraints, including soft-real time constraints. This will create the need for a computing infrastructure where data parallel applications can be run. Data distribution for such applications assumes an important role for the performance of the system. In this paper, we address the problem of data partitioning as part of data distribution, such that parallel analysis of the partitions can minimize the overall run-time of the analysis. We investigate the problem under the scenarios where the communication links and computation nodes are unreliable and intermittently unavailable. We assume all such non-availabilities are known to the partitioner and propose an algorithm which generates a static partition of the data based on the capacity and availability of the elements in the system.