Control theory for principled heap sizing

White, David R.; Singer, Jeremy; Aitken, Jonathan M.; Jones, Richard

doi:10.1145/2491894.2466481

Cited by 14 publications

(7 citation statements)

References 34 publications

(19 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…They conclude that in most of the cases, the Heap Availability Policy yields the best results. Finally, White et al [15] propose a strategy for adaptive resizing of the whole heap based on control theory. They show that heap sizing can be formulated as a control problem and that a standard controller program can be applied, yielding good improvements for some benchmarks.…”

Section: Related Workmentioning

confidence: 99%

Kindergarten cop: dynamic nursery resizing for GHC

Ferreiro

Castro

Janjić

et al. 2016

Proceedings of the 25th International Conference on Compiler Construction

View full text Add to dashboard Cite

Generational garbage collectors are among the most popular garbage collectors used in programming language runtime systems. Their performance is known to depend heavily on choosing the appropriate size of the area where new objects are allocated (the nursery). In imperative languages, it is usual to make the nursery as large as possible, within the limits imposed by the heap size. Functional languages, however, have quite different memory behaviour. In this paper, we study the effect that the nursery size has on the performance of lazy functional programs, through the interplay between cache locality and the frequency of collections. We demonstrate that, in contrast with imperative programs, having large nurseries is not always the best solution. Based on these results, we propose two novel algorithms for dynamic nursery resizing that aim to achieve a compromise between good cache locality and the frequency of garbage collections. We present an implementation of these algorithms in the state-of-the-art GHC compiler for the functional language Haskell, and evaluate them using an extensive benchmark suite. In the best case, we demonstrate a reduction in total execution times of up to 88.5%, or an 8.7 overall speedup, compared to using the production GHC garbage collector. On average, our technique gives an improvement of 9.3% in overall performance across a standard suite of 63 benchmarks for the production GHC compiler.

show abstract

Section: Related Workmentioning

confidence: 99%

Kindergarten cop: dynamic nursery resizing for GHC

Ferreiro

Castro

Janjić

et al. 2016

Proceedings of the 25th International Conference on Compiler Construction

View full text Add to dashboard Cite

show abstract

“…At the moment, the system uses the microeconomic optimization model outlined in Section 3, but this could be replaced by another model, such as the market-based approach proposed by Vengerov [21], a control system [23] or a search-based approach. The daemon could be extended with a plugin system to allow these optimization models to be replaced and modified by a system administrator.…”

Section: Future Workmentioning

confidence: 99%

The judgment of forseti: economic utility for dynamic heap sizing of multiple runtimes

Cameron

Singer

Vengerov

2015

Proceedings of the 2015 International Symposium on Memory Management

Self Cite

View full text Add to dashboard Cite

We introduce the Forseti system, which is a principled approach for holistic memory management. It permits a sysadmin to specify the total physical memory resource that may be shared between all concurrent virtual machines on a physical node. Forseti models the heap size versus application throughput for each virtual machine, and seeks to maximize the combined throughput of the set of VMs based on concepts from economic utility theory. We evaluate the Forseti system using a standard Java managed runtime, i.e. OpenJDK. Our results demonstrate that Forseti enables dramatic reductions (up to 5x) in heap footprint without compromising application execution times.

show abstract

“…These data can include total memory size, heap size, GC efficiency, current memory footprint, real memory usage, memory pressure, empirically collected data about program behaviour, etc. Some of the algorithms rely on empirically tuned values [12,13] or on data collected by running benchmarks [11,14]. Other algorithms require changes in VM or OS or both [15].…”

Section: Mark-bits Organizationmentioning

confidence: 99%

“…Heap sizing policy is often hard-coded in a virtual machine and based on trial-and-error experiments [14]. In this research we developed heap expansion policy based on objects' behaviour rather than on values such as time taken for GC or current amount of live data in the heap.…”

Section: Heuristic Heap Expansionmentioning

confidence: 99%

Mark without much Sweep Algorithm for Garbage Collection

2014

View full text Add to dashboard Cite

In this paper two simple improvements over traditional mark-sweep collector are proposed. The core idea is placing small objects of the same type in buckets. The buckets are organised in such way to eliminate the internal fragmentation, sweeping, and freeing inside them. The measured improvement of garbage collection time over traditional mark-sweep is 19%. Another proposed improvement is more general and is applicable to other garbage collection algorithms as well. It uses heuristics to control the heap growth. The regularities in behaviour of objects of particular types are used to determine whether the collection should be performed or avoided in favour of immediate heap expansion. The heap expansion algorithm reduces garbage collection time over traditional marksweep for 49% while keeping the heap size approximately the same. Key words: Mark-sweep, Garbage collection, Heap sizing, Memory managementAlgoritam za zbrinjavanje memorije s označavanjem i smanjenim oslobadanjem. U ovomčlanku opisana su dva jednostavna poboljšanja algoritma označi-oslobodi. Osnovna ideja jest smještanje malih objekata istog tipa u pretince. Pretinci su organizirani tako da se u njima ne pojavljuje unutarnja fragmentacija, a uklanja se i potreba za oslobadanjem blokova zauzetih nedohvatljivim objektima. Vrijeme provedeno u zbrinjavanju manje je za 19% u odnosu na klasični algoritam označi-oslobodi. Drugo poboljšanje je općenitije i moguće ga je primijeniti i na druge algoritme za zbrinjavanje memorije. U njemu rastom gomile upravlja heuristički algoritam koji koristi pravilnosti u ponašanju objekata različitih tipova. Na temelju njih, algoritam odlučuje hoće li gomila biti zbrinuta ili odmah proširena. Heuristička inačica algoritma smanjuje vrijeme provedeno u zbrinjavanju u odnosu na tradicionalni algoritam označi-oslobodi za 49%, a da pri tome zahtijeva približno istu količinu memorije.

show abstract

Control theory for principled heap sizing

Cited by 14 publications

References 34 publications

Kindergarten cop: dynamic nursery resizing for GHC

Kindergarten cop: dynamic nursery resizing for GHC

The judgment of forseti: economic utility for dynamic heap sizing of multiple runtimes

Mark without much Sweep Algorithm for Garbage Collection

Contact Info

Product

Resources

About