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

»çÀÌÆ®¸Ê

Loading..

Please wait....

±¹³» ÇÐȸÁö

Ȩ Ȩ > ¿¬±¸¹®Çå > ±¹³» ÇÐȸÁö > µ¥ÀÌÅͺ£À̽º ¿¬±¸È¸Áö(SIGDB)

µ¥ÀÌÅͺ£À̽º ¿¬±¸È¸Áö(SIGDB)

Current Result Document :

ÇѱÛÁ¦¸ñ(Korean Title) ¸ÖƼÄÚ¾î ȯ°æÀÇ In-Memory µ¥ÀÌÅͺ£À̽º »ó¿¡¼­ È¿À²ÀûÀÎ Æ®·£Àè¼Ç 󸮸¦ À§ÇÑ Ãæµ¹ ¿¹Ãø ±â¹Ý Æ®·£Àè¼Å³Î ¸Þ¸ð¸® ±â¹ý
¿µ¹®Á¦¸ñ(English Title) Conflict Prediction based Transactional Memory Technique for Efficient Transaction Processing in Multi-core In-Memory Databases
ÀúÀÚ(Author) À±¹Î   °­¹®È¯   À忬¿ì   ÀåÀç¿ì   Yoon Min   Mun-Hwan Kang   Yeon-Woo Jang   Jae-Woo Chang  
¿ø¹®¼ö·Ïó(Citation) VOL 33 NO. 02 PP. 0163 ~ 0179 (2017. 08)
Çѱ۳»¿ë
(Korean Abstract)
ÃÖ±Ù Intel¿¡¼­ óÀ½À¸·Î »ó¿ëÈ­µÈ Hardware Transactional Memory(HTM)·Î½á Transactional Synchronization Extension(TSX)¸¦ Á¦¾ÈÇÏ¿´´Ù. HTMÀº Æ®·£Àè¼Ç 󸮸¦ À§ÇÑ º´·Ä ÇÁ·Î±×·¡¹Ö Æз¯´ÙÀÓÀ»Å©°Ô ¹Ù²Ù¾úÀ¸¸ç, À̸¦ ÀÌ¿ëÇÑ ´Ù¼öÀÇ ¿¬±¸µéÀÌ Á¦¾ÈµÇ¾ú´Ù. ±×·¯³ª, ±âÁ¸ ¿¬±¸µéÀº ´Ù¾çÇÑ ¿öÅ©·Îµå¿¡ ´ëÇØ È¹ÀÏÈ­µÈ ȯ°æÀ» Á¦°øÇÏ´Â ¹®Á¦Á¡ÀÌ Á¸ÀçÇÑ´Ù. µû¶ó¼­ ÀÌ·¯ÇÑ ¹®Á¦Á¡À» ÇØ°áÇϱâ À§ÇØ, º» ³í¹®¿¡¼­´Â ¸ÖƼÄÚ¾î Àθ޸𸮠ȯ°æ¿¡¼­ÀÇ È¿À²ÀûÀÎ Æ®·£Àè¼Ç 󸮸¦ À§ÇÑ Ãæµ¹ ¿¹Ãø ±â¹Ý Æ®·£Àè¼Å³Î ¸Þ¸ð¸® ±â¹ýÀ» Á¦¾ÈÇÑ´Ù. Á¦¾ÈÇÏ´Â Ãæµ¹ ¿¹Ãø ±â¹Ý Æ®·£Àè¼Å³Î ¸Þ¸ð¸® ±â¹ýÀº Ãæµ¹ ¿¹ÃøÀ» À§ÇÑ ¸ÅÆ®¸¯½º¸¦ ±¸¼ºÇϸç, °¢ Æ®·£Àè¼Ç¿¡ ´ëÇÑ ¸ÞŸµ¥ÀÌÅ͸¦ ÃàÀûÇÏ°í, ´Ù¾çÇÑ Æ®·£Àè¼ÇÀÇ abort ºÐ¼®À» ÅëÇÑ Ãæµ¹ ¿¹ÃøÀ» Á¦°øÇÑ´Ù. À̸¦ ÅëÇØ Æ®·£Àè¼ÇÀÇ Ãæµ¹ ¿¹ÃøÀ» ¼öÇàÇÏ¿© ÃÖ»óÀÇ ¼º´ÉÀ» º¸ÀåÇÏ´Â HTM ¹× Á÷·Ä ½ÇÇàÀ» ¼öÇàÇÔÀ¸·Î½á, ¸ÖƼ ÄÚ¾î ȯ°æ¿¡¼­ Æ®·£Àè¼Ç ó¸® È¿À²À» Çâ»ó½ÃŲ´Ù. ¾Æ¿ï·¯, °æ»çÇÏ°­¹ý ±â¹Ý jump ¾Ë°í¸®Áò ¹× Æ®·£Àè¼Ç ÇÊÅ͸µ ±â¹ýÀ» ÅëÇØ °¢ ¿öÅ©·Îµå ³»¿¡¼­ Æ®·£Àè¼ÇÀÇ Æ¯¼ºÀ» ¹Ý¿µÇÑ È¿À²ÀûÀÎ HTM Àç½Ãµµ Á¤Ã¥À» Á¦°øÇÑ´Ù. ¸¶Áö¸·À¸·Î STAMP¸¦ ÅëÇÑ ¼º´ÉÆò°¡¸¦ ÅëÇØ, Á¦¾ÈÇÏ´Â ±â¹ýÀÌ ±âÁ¸ ±â¹ý¿¡ ºñÇØ 50-200%ÀÇ ¼º´É Çâ»óÀ» ´Þ¼ºÇÔÀ» º¸ÀδÙ.
¿µ¹®³»¿ë
(English Abstract)
Recently, Intel proposed Transactional Synchronization Extension (TSX) as its first mainstream Hardware Transactional Memory (HTM), HTM has greatly changed the parallel programming paradigm for transaction processing. As a result, a number of studies on HTM have been actively done. However, the existing studies have a problem that they provide a static HTM configuration for all workloads. To solve the problem, we propose a hardware transactional memory scheme based on conflict prediction for efficient transaction processing in multi-core in-memory environments. First, the proposed HTM scheme constructs a matrix for conflict prediction, collects metadata about transactions, and provide conflict prediction through the analysis of various transactions aborts. As a result, the proposed HTM scheme improves the efficiency of transaction processing in multi-core in-memory environments by performing the conflict prediction of transactions. In addition, the proposed scheme provides an efficient HTM retry policy according to the characteristic of a given workload by using a jump algorithm based on gradient descent exploration. Finally, it is shown through our performance analysis using STAMP that the proposed scheme achieves about 50~200% better performance than the existing HTM schemes.
Å°¿öµå(Keyword) Çϵå¿þ¾î Æ®·£Àè¼Å³Î ¸Þ¸ð¸®   Ãæµ¹ ¿¹Ãø   Àç½Ãµµ Á¤Ã¥   ¸ÖƼÄÚ¾î Àθ޸𸮠ȯ°æ   Hardware Transactional Memory   conflict prediction   retry policy   multi-core in-memory database  
ÆÄÀÏ÷ºÎ PDF ´Ù¿î·Îµå