"Stream processing applications process high-volume, continuous feeds from live data sources, employ data-in-motion analytics to analyze these feeds, and produce near real-time results with low latency. With the explosion in the amount of data available as live feeds, stream computing has found wide applications in areas ranging from telecommunications to health-care to cyber-security. The high volume of the data to be processed, the velocity at which the results need to be produced, and the variety of the data sources involved make stream processing applications unique and challenging. One of the major limitations of existing stream processing middleware is their inability to take advantage of parallelism opportunities that exist in stream processing applications, in order to scale in the presence of additional resources and workload. This often requires manual intervention by the application developers to specify what is 'safe' to parallelize and/or how much parallelization is 'profitable'. Yet another challenge is to dynamically adapt to the changes in workload and resource availability. We propose system-level techniques that address these challenges via 'transparent' and 'elastic' scaling. The transparent nature of the scaling means that the application developers do not need to specify what parts of the application is safe to parallelize, and instead this information will be derived through automatic analysis.The elastic nature of the scaling means that the level of parallelism is dynamically adjusted as resource and workload availability changes. Furthermore, in the presence of multiple kinds of parallelism as well as multiple instances of application segments that can benefit from these, we propose techniques to effectively manage resources in order to maximize application performance."