Bottle graphs: visualizing scalability bottlenecks in multi-threaded applications

Understanding and analyzing multi-threaded program performance and scalability is far from trivial, which severely complicates parallel software development and optimization. In this paper, we present bottle graphs, a powerful analysis tool that visualizes multi-threaded program performance, in regards to both per-thread parallelism and execution time. Each thread is represented as a box, with its height equal to the share of that thread in the total program execution time, its width equal to its parallelism, and its area equal to its total running time. The boxes of all threads are stacked upon each other, leading to a stack with height equal to the total program execution time. Bottle graphs show exactly how scalable each thread is, and thus guide optimization towards those threads that have a smaller parallel component (narrower), and a larger share of the total execution time (taller), i.e. to the 'neck' of the bottle. Using light-weight OS modules, we calculate bottle graphs for unmodified multi-threaded programs running on real processors with an average overhead of 0.68%. To demonstrate their utility, we do an extensive analysis of 12 Java benchmarks running on top of the Jikes JVM, which introduces many JVM service threads. We not only reveal and explain scalability limitations of several well-known Java benchmarks; we also analyze the reasons why the garbage collector itself does not scale, and in fact performs optimally with two collector threads for all benchmarks, regardless of the number of application threads. Finally, we compare the scalability of Jikes versus the OpenJDK JVM. We demonstrate how useful and intuitive bottle graphs are as a tool to analyze scalability and help optimize multi-threaded applications.Understanding and analyzing multi-threaded program performance and scalability is far from trivial, which severely complicates parallel software development and optimization. In this paper, we present bottle graphs, a powerful analysis tool that visualizes multi-threaded program performance, in regards to both per-thread parallelism and execution time. Each thread is represented as a box, with its height equal to the share of that thread in the total program execution time, its width equal to its parallelism, and its area equal to its total running time. The boxes of all threads are stacked upon each other, leading to a stack with height equal to the total program execution time. Bottle graphs show exactly how scalable each thread is, and thus guide optimization towards those threads that have a smaller parallel component (narrower), and a larger share of the total execution time (taller), i.e. to the 'neck' of the bottle. Using light-weight OS modules, we calculate bottle graphs for unmodified multi-threaded programs running on real processors with an average overhead of 0.68%. To demonstrate their utility, we do an extensive analysis of 12 Java benchmarks running on top of the Jikes JVM, which introduces many JVM service threads. We not only reveal and explain scalability limitations of several well-known Java benchmarks; we also analyze the reasons why the garbage collector itself does not scale, and in fact performs optimally with two collector threads for all benchmarks, regardless of the number of application threads. Finally, we compare the scalability of Jikes versus the OpenJDK JVM. We demonstrate how useful and intuitive bottle graphs are as a tool to analyze scalability and help optimize multi-threaded applications.A