Garbage collection offers numerous software engineering advantages, but interacts poorly with virtual memory managers. Existing garbage collectors require far more pages than the application’s working set and touch pages without regard to which ones are in memory, especially during full-heap garbage collection. The resulting paging can cause throughput to plummet and pause times to spike up to seconds or even minutes.
I present a garbage collector that avoids paging. This bookmarking collector cooperates with the virtual memory manager to guide its eviction decisions. Using summary information (“bookmarks”) recorded from evicted pages, the collector can perform in-memory full-heap collections. In the absence of memory pressure, the bookmarking collector matches the throughput of the best collector we tested while running in smaller heaps. In the face of memory pressure, it improves throughput by up to a factor of five and reduces pause times by up to a factor of 45 over the next best collector. Compared to a collector that consistently provides high throughput (generational mark-sweep), the bookmarking collector reduces pause times by up to 218x and improves throughput by up to 41x. Bookmarking collection thus provides greater utilization of available physical memory than other collectors while matching or exceeding their throughput.
This is joint work with Matthew Hertz and Yi Feng at the University of Massachusetts Amherst.