[HAL] BigDataFr recommends: NumaGiC – a Garbage Collector for Big Data on Big NUMA Machines #datascientist #machine learning

BigDataFr recommends: NumaGiC – a Garbage Collector for Big Data on Big NUMA Machines

Introduction (excerpt)

‘Data-analytics programs require large amounts of computing power  and  memory.  When  run  on  modern  multicore  computers with a cache-coherent Non-Uniform Memory Access (ccNUMA)  architecture,  they  suffer  from  a  high  overhead during garbage collection (GC) caused by a bad memory access locality. A ccNUMA architecture consists of a network of nodes, each comprising several cores and a local memory bank.

A ccNUMA architecture hides the distributed nature of the memory from the application. The application thus unknowingly creates inter-node references when it stores a reference to an object located on a given node into the memory of another node.’ […]

Read paper
By Lokesh Gidra, LIP6-UPMC/INRIA; Gaël Thomas, SAMOVAR-Telecom SudParis; Julien Sopena, LIP6-UPMC/INRIA; Marc Shapiro, LIP6-INRIA/UPMC; Nhan Nguyen, Chalmers University of Technology
Source: hal.archives-ouvertes.fr

Laisser un commentaire

Votre adresse e-mail ne sera pas publiée. Les champs obligatoires sont indiqués avec *