Time-triggered garbage collection

Abstract

The advent of Java and similar languages on the real-time system scene necessitates the development of efficient strategies for scheduling the work of a garbage collector in a non-intrusive way. We propose a scheduling strategy, timetriggered garbage collection, based on assigning the collector a deadline for when it must complete its current cycle.We show that a time-triggered GC with fixed deadline can have equal or better real-time performance than an allocation-triggered GC, which is the standard approach to real-time GC. Also, by using a deadline-based approach, the GC scheduling and, consequently, real-time performance, is independent of a complex and error-prone GC work metric.Time-triggered GC allows a more high-level view on GC scheduling; we look at the GC cycle level rather than at the individual work increments. This makes it possible to schedule GC as any other thread. It is also suitable for making the GC auto-tuning by dynamically adjusting its deadline as necessary.We have implemented our approach in a run-time system for Java and present experimental data to support the practical feasibility of the approach.

The space cost of lazy reference counting

“…The large blocks corresponding to an empty list are immediately dropped. 17 Note that an immediately reclaimable small object in P programmable may correspond to an immediately reclaimable large object in P proo f , but neither will be freed and hence, by our assumptions, the difference will not affect the deallocation sequence.…”

“…This is closely related to making a reference-count implementation usable in real-time applications, though there seems to be some disagreement on the precise requirements for real-time memory management (cf. [2,17]). Certainly a reference count implementation cannot guarantee a minimum processor utilization for a given section of client code unless it is possible to bound the number of free calls made on its behalf from the reference count implementation during that section of code.…”

“…16 Since none of the large objects P experiment are ever freed, inductively P experiment and P proo f generate isomorphic free graphs after each main loop iteration. 17 Specifically, P proo f must still free exactly k i elements from the i th list, and P proo f still fails to free any of the large objects.…”

“…The maximum pause times were obtained allocating and dropping height 16 trees, i.e. trees containing 2 17 1 vertices. (It also includes some initialization code that allocates a single tree occupying the entire heap to ensure a reasonable heap size.…”