30.01.2014 Aufrufe

Abstract-Band - Fakultät für Informatik, TU Wien - Technische ...

Abstract-Band - Fakultät für Informatik, TU Wien - Technische ...

Abstract-Band - Fakultät für Informatik, TU Wien - Technische ...

MEHR ANZEIGEN
WENIGER ANZEIGEN

Erfolgreiche ePaper selbst erstellen

Machen Sie aus Ihren PDF Publikationen ein blätterbares Flipbook mit unserer einzigartigen Google optimierten e-Paper Software.

Michael Starzinger<br />

Exact Garbage Collection for the Cacao Virtual Machine<br />

Studium: Masterstudium Software Engineering & Internet Computing<br />

Betreuer: Ao.Univ.Prof. Dr. Andreas Krall<br />

Every virtual machine capable of executing Java byte-code needs to perform<br />

some sort of automatic memory management, commonly referred to as<br />

garbage collection. However, it is not specified how this garbage collection<br />

process actually has to be performed. So far the Cacao Virtual Machine<br />

(CacaoVM) used a conservative garbage collector that was designed for<br />

uncooperative environments, thus it has been unable to take full advantage<br />

of the information already present in the runtime infrastructure. This<br />

conservative garbage collector was replaced by an exact one, which has been<br />

tailored to the specific needs of a Java Runtime Environment. In the context<br />

of Java several different components of the virtual machine need to be<br />

orchestrated for exact garbage collection to work. The application threads act<br />

as mutators on the heap and need to be prevented from interfering with the<br />

collector. Another central aspect is identifying the actual references which<br />

point into the heap and prevent objects from being collected. These<br />

references are referred to as the root-set and keep objects alive. The actual<br />

collection algorithms are in charge of cleaning up dead objects and<br />

reorganizing the heap, so that free memory is made available again. Those<br />

algorithms have already been researched in depth over the past decades,<br />

exhibiting vastly different characteristics. All of the above aspects need to be<br />

considered from the perspective of a pure Just-In-Time compilation approach<br />

as taken by CacaoVM. Some crucial simplifications, which rely on fallbacks<br />

into an unoptimized interpreter mode, cannot be applied and need to be<br />

revisited in this setting. Special care was taken to allow future development<br />

towards a generational garbage collection approach by providing the<br />

necessary infrastructure.<br />

Michael Vögler<br />

SILCA - Self-Initiative Load Clustering Agents<br />

Studium: Masterstudium Software Engineering & Internet Computing<br />

Betreuerin: Ao.Univ.Prof. Dr. Eva Kühn<br />

58<br />

To handle the load of a service more efficiently load clustering can be used to<br />

cluster a set of loads into smaller subsets, where each of this so called clusters<br />

captures a specific aspect of the load. There exist many clustering algorithms<br />

reaching from unintelligent to intelligent ones. However most approaches are<br />

very problem oriented and therefore hard to compare. In this thesis we<br />

propose a generic architectural pattern for a load clustering framework that<br />

allows the plugging of different clustering and classification algorithms.<br />

Furthermore this pattern should ease the selection of the best algorithm for a<br />

certain problem scenario. The presented pattern assumes autonomous

Hurra! Ihre Datei wurde hochgeladen und ist bereit für die Veröffentlichung.

Erfolgreich gespeichert!

Leider ist etwas schief gelaufen!