• Àüü
  • ÀüÀÚ/Àü±â
  • Åë½Å
  • ÄÄÇ»ÅÍ
´Ý±â

»çÀÌÆ®¸Ê

Loading..

Please wait....

±¹³» ³í¹®Áö

Ȩ Ȩ > ¿¬±¸¹®Çå > ±¹³» ³í¹®Áö > Çѱ¹Á¤º¸°úÇÐȸ ³í¹®Áö > Á¤º¸°úÇÐȸ ÄÄÇ»ÆÃÀÇ ½ÇÁ¦ ³í¹®Áö (KIISE Transactions on Computing Practices)

Á¤º¸°úÇÐȸ ÄÄÇ»ÆÃÀÇ ½ÇÁ¦ ³í¹®Áö (KIISE Transactions on Computing Practices)

Current Result Document : 11 / 11

ÇѱÛÁ¦¸ñ(Korean Title) ¸ÖƼ Äھ À§ÇÑ È¿À²ÀûÀÎ JAVA Full GC Compact ¹æ¹ý
¿µ¹®Á¦¸ñ(English Title) Efficient Java Full GC Compact Scheme for Multicores
ÀúÀÚ(Author) ¹Îö±â   ¿°Ç念   ÇÑ Çõ   Cheol Gi Min   Heon Y. Yeom   Hyuck Han  
¿ø¹®¼ö·Ïó(Citation) VOL 27 NO. 03 PP. 0129 ~ 0136 (2021. 03)
Çѱ۳»¿ë
(Korean Abstract)
º» ¿¬±¸¿¡¼­´Â ±âÁ¸ÀÇ JAVA Full GC Parallel Compact¸¦ ´ëüÇÒ ¼ö ÀÖ´Â »õ·Î¿î compact ¹æ¹ýÀ» Á¦½ÃÇÑ´Ù. ±âÁ¸ ¹æ¹ýÀ» ºÐ¼®ÇÏ°í ¹®Á¦Á¡À» ã¾Æ³»¸ç À̸¦ ÇØ°áÇϱâ À§ÇØ compact¸¦ º´·Ä ó¸®ÇÏ´Â ºÎºÐ°ú ´Üµ¶ ó¸®ÇÏ´Â ºÎºÐÀ¸·Î ºÐ¸®ÇÑ´Ù. º´·Ä ó¸® ºÎºÐ¿¡¼­´Â °¢ ¿µ¿ª(region) º°·Î ¿µ¿ª ³»ºÎÀÇ °´Ã¼µé À» µ¶¸³ÀûÀ¸·Î ó¸®Çϵµ·Ï ÇÏ¿© º´·Ä¼ºÀ» ±Ø´ëÈ­ÇÑ´Ù. ±× µÚ¿¡ ´Üµ¶ ó¸® ºÎºÐ¿¡¼­ °³º°ÀûÀ¸·Î ó¸®µÈ ¿µ¿ªµéÀ» ¿µ¿ª ¾ÕÂÊÀ¸·Î ¸ð¾Æ ±âÁ¸ÀÇ compact¿Í µ¿ÀÏÇÑ °á°ú¹°À» ¸¸µé°Ô ÇÑ´Ù. ÀÌ ¶§, ¸®´ª½º Ä¿³Î¿¡ ±¸ÇöÇÑ ÆäÀÌÁö Å×À̺í Á¶ÀÛ syscall ÇÔ¼ö¸¦ ÀÌ¿ëÇÏ¿© ÇϳªÀÇ ¾²·¹µå·Îµµ È¿À²ÀûÀΠ󸮰¡ °¡´ÉÇÏ°Ô ÇÑ´Ù. Æò°¡¸¦ À§ÇØ dacapo, specjvm2008 º¥Ä¡¸¶Å©¿Í Apache Spark »ó¿¡¼­ ±¸ÇöÇÑ PageRank¸¦ ÅëÇØ ½ÇÇèÇÏ¿© Á¦¾ÈÇÑ ¹æ¹ýÀÇ ¿ì¼ö¼ºÀ» º¸ÀÌ°íÀÚ ÇÑ´Ù.
¿µ¹®³»¿ë
(English Abstract)
In this study, we propose a new compact algorithm that can replace existing the JAVA Full GC Parallel Compact. We first analyze the existing compact algorithm and characterize the inefficiencies of the algorithm. Then, to address these inefficiencies, we approach the problem into two parts by splitting the compact job into multi-threaded work and single-threaded work. In the multi-threaded part, we maximize parallelism by independently processing objects in regions. Then, in the single-threaded part, we move the regions to the start of space and obtain the same result with existing algorithm. In this process, we use page table manipulation for efficient region movement. Finally, we evaluate our scheme using dacapo, SPECjvm2008 benchmark, and PageRank on top of Apache Spark.
Å°¿öµå(Keyword) ¾²·¹±â ¼öÁý   ÀÚ¹Ù °¡»ó ¸Ó½Å   ¸ÖƼÄھ º´·Ä 󸮠  garbage collection   java virtual machine   multicore   parallel processing  
ÆÄÀÏ÷ºÎ PDF ´Ù¿î·Îµå