Large-scale multicore architectures create new challenges for garbage collectors (GCs). In particular, throughput-oriented stop-the-world algorithms demonstrate good performance with a small number of cores, but have been shown to degrade badly beyond approximately 8 cores on 48-core OpenJDK 7. This negative result raises the question whether design has intrinsic limitations that would require radically different approach. Our study suggests answer is no, and there no compelling scalability...

On contemporary cache-coherent Non-Uniform Memory Access (ccNUMA) architectures, applications with a large memory footprint suffer from the cost of garbage collector (GC), because, as GC scans reference graph, it makes many remote accesses, saturating interconnect between nodes. We address this problem NumaGiC, mostly-distributed design. In order to maximise access locality during collection, thread avoids accessing different node, instead notifying message; nonetheless, NumaGiC drawbacks...

Managed Runtime Environments (MRE) are increasingly used for application servers that use large multi-core hardware. We find the garbage collector is critical overall performance in this setting. explore costs and scalability of collectors on a contemporary 48-core multiprocessor machine. present experimental evaluation parallel concurrent OpenJDK, widely-used Java virtual show collection represents substantial amount an application's execution time, does not scale well as number cores...

Managed Runtime Environments (MRE) are increasingly used for application servers that use large multi-core hardware. We find the garbage collector is critical overall performance in this setting. explore costs and scalability of collectors on a contemporary 48-core multiprocessor machine. present experimental evaluation parallel concurrent OpenJDK, widely-used Java virtual show collection represents substantial amount an application's execution time, does not scale well as number cores...

Large-scale multicore architectures create new challenges for garbage collectors (GCs). In particular, throughput-oriented stop-the-world algorithms demonstrate good performance with a small number of cores, but have been shown to degrade badly beyond approximately 8 cores on 48-core OpenJDK 7. This negative result raises the question whether design has intrinsic limitations that would require radically different approach. Our study suggests answer is no, and there no compelling scalability...

The performance of mobile devices directly affects billions people worldwide. Yet, despite memory management being key to their responsiveness, energy efficiency, and cost, are understudied in the literature. A paucity suitable methodologies benchmarks is likely both a cause consequence this gap. It also reflects challenges evaluating due to: i) inherently multi-tenanted nature, ii) scarcity widely-used open source workloads as benchmarks, iii) challenge determinism reproducibility given...

On contemporary cache-coherent Non-Uniform Memory Access (ccNUMA) architectures, applications with a large memory footprint suffer from the cost of garbage collector (GC), because, as GC scans reference graph, it makes many remote accesses, saturating interconnect between nodes. We address this problem NumaGiC, mostly-distributed design. In order to maximise access locality during collection, thread avoids accessing different node, instead notifying message; nonetheless, NumaGiC drawbacks...

On contemporary cache-coherent Non-Uniform Memory Access (ccNUMA) architectures, applications with a large memory footprint suffer from the cost of garbage collector (GC), because, as GC scans reference graph, it makes many remote accesses, saturating interconnect between nodes. We address this problem NumaGiC, mostly-distributed design. In order to maximise access locality during collection, thread avoids accessing different node, instead notifying message; nonetheless, NumaGiC drawbacks...

Large-scale multicore architectures create new challenges for garbage collectors (GCs). In particular, throughput-oriented stop-the-world algorithms demonstrate good performance with a small number of cores, but have been shown to degrade badly beyond approximately 8 cores on 48-core OpenJDK 7. This negative result raises the question whether design has intrinsic limitations that would require radically different approach. Our study suggests answer is no, and there no compelling scalability...