• 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 : 9 / 10 ÀÌÀü°Ç ÀÌÀü°Ç   ´ÙÀ½°Ç ´ÙÀ½°Ç

ÇѱÛÁ¦¸ñ(Korean Title) ¿­±â¹Ý DBMS¸¦ À§ÇÑ SIMD ¸í·É¾î ±â¹ÝÀÇ È¿À²ÀûÀÎ °Ë»ö ¿¬»êÀÚ
¿µ¹®Á¦¸ñ(English Title) An Efficient Scan Operator Based on SIMD Instructions for Column Oriented DBMSs
ÀúÀÚ(Author) ÃÖ¼ºÁØ   ¹ÎÁر⠠ Seongjun Choe   Jun-Ki Min  
¿ø¹®¼ö·Ïó(Citation) VOL 33 NO. 03 PP. 0089 ~ 0099 (2017. 12)
Çѱ۳»¿ë
(Korean Abstract)
°ü°è ¿¬»ê¿¡¼­ È¿À²ÀûÀÎ µ¥ÀÌÅÍ Ã³¸®¸¦ À§ÇØ SIMD (Single Instruction Multiple Data)¸¦ È°¿ëÇÏ´Â µ¥ÀÌÅÍ Ã³¸® ±â¹ý¿¡ ´ëÇÏ¿© ¿¬±¸°¡ ÁøÇàµÇ¾î ¿Ô´Ù. SIMD´Â SIMD ·¹Áö½ºÅÍ¿¡ º¹¼ö °³ÀÇ µ¥ÀÌÅ͸¦ ÀûÀçÇÏ°í SIMD ¸í·É¾î·Î µ¿½Ã¿¡ ó¸®ÇÏ´Â ´ëÇ¥ÀûÀÎ º´·ÄÈ­ ¹æ¹ýÀÌ´Ù. °Ë»ö ¿¬»êÀº µ¥ÀÌÅͺ£À̽º¿¡¼­ ±âº»ÀûÀÎ °ü°è ¿¬»êÀ¸·Î ÁúÀÇ¿¡¼­ ÁÖ¾îÁø Á¶°ÇÀ» ¸¸Á·ÇÏ´Â µ¥ÀÌÅ͸¸À» °Ë»öÇÏ´Â ¿¬»êÀÌ´Ù. º» ³í¹®¿¡¼­´Â ºÐ±â¸¦ ÃÖ¼ÒÈ­ÇÏ¿© ºÐ±â ¿¹Ãø ½ÇÆзΠÀÎÇÑ ¼º´É ÀúÇÏÀÇ À¯¹ßÀ» ÃÖ¼ÒÈ­ÇÏ°í µ¥ÀÌÅÍ º´·ÄÈ­¸¦ ÀÌ·ç±â À§ÇØ SIMDÀÇ ¸í·É¾îµéÀ» È°¿ëÇÑ SIMD ±â¹ÝÀÇ °Ë»öÀ» Á¦½ÃÇÑ´Ù. ´õºÒ¾î, Á¦¾ÈÇÑ SIMD ±â¹ÝÀÇ °Ë»öÀ» »ó¿ë ¿­ ±â¹Ý DBMS (Database Management System)ÀÎ MonetDB¿¡ Àû¿ëÇÏ°í ½ÇÇèÀ» ÅëÇØ º» ³í¹®¿¡¼­ Á¦¾ÈÇÏ´Â SIMD ±â¹ÝÀÇ °Ë»öÀÌ È¿À²ÀûÀÓÀ» º¸¿´´Ù.
¿µ¹®³»¿ë
(English Abstract)
Data processinng techniques utilizing SIMD (Single Instruction Multiple Data) for efficient data processing in relational operations have been studied. SIMD is a representative parallelization method that loads multiple data into a SIMD register and process them simultaneously with SIMD instructions. As a basic relational operation in the database, scan operation selects the data satisfying a given query condition. In this paper, we propose a SIMD-based scan by using minimal branch condition resulting in reducing the performance degradation due to misprediction and by utilizing SIMD instructions to achieve data parallelism. In addition, we implement our proposed SIMD-based scan on a commercial column-oriented DBMS MonetDB and show the efficiency of our proposed SIMD-based scan.
Å°¿öµå(Keyword) Database   SIMD   MonetDB  
ÆÄÀÏ÷ºÎ PDF ´Ù¿î·Îµå