• 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 ³í¹®Áö D : µ¥ÀÌŸº£À̽º

Á¤º¸°úÇÐȸ ³í¹®Áö D : µ¥ÀÌŸº£À̽º

Current Result Document :

ÇѱÛÁ¦¸ñ(Korean Title) Ç÷¡½Ã SSD¿¡¼­ È¿À²ÀûÀÎ À妽º ½ºÄµÀ» À§ÇÑ ºñµ¿±â ´ÙÁß ÆäÀÌÁö Àбâ
¿µ¹®Á¦¸ñ(English Title) Asynchronous Multiple Page Reading for Efficient Index Scans on Flash SSDs
ÀúÀÚ(Author) ¹ÚÁö¿µ   °­¿îÇР  ÀÌ»ó¿ø   Ji-Young Park   Woon-Hak Kang   Sang-Won Lee  
¿ø¹®¼ö·Ïó(Citation) VOL 41 NO. 01 PP. 0037 ~ 0044 (2014. 02)
Çѱ۳»¿ë
(Korean Abstract)
ÇöÀçÀÇ ¸ðµç DBMS°¡ äÅÃÇÏ°í ÀÖ´Â À妽º ½ºÄµ ¹æ½Ä¿¡¼­´Â À妽º ¸®ÇÁ³ëµåÀÇ À妽º ¿£Æ®¸®°¡ °¡¸®Å°´Â µ¥ÀÌÅÍ ÆäÀÌÁöµéÀÌ µ¿±â Àб⠿äû ¹æ½ÄÀ¸·Î µ¥ÀÌÅͺ£À̽º ¹öÆÛ¿¡ ÀÐÇôÁø´Ù. µû¶ó¼­, °Ë»ö Á¶°ÇÀ» ¸¸Á·ÇÏ´Â ·¹ÄÚµåµéÀÌ ¿©·¯ µ¥ÀÌÅÍ ÆäÀÌÁö¿¡ Èð¾îÁ® ÀÖ´Â ºñ±ºÁý À妽ºÀÇ °æ¿ì, ÇöÀçÀÇ À妽º ½ºÄµ ¹æ½ÄÀº ÁúÀÇ Ã³¸® ½Ã Ç÷¡½Ã SSDÀÇ ³»ºÎ º´·Ä¼ºÀ» È°¿ëÇÏÁö ¸øÇÏ°í, ¶ÇÇÑ °úµµÇÑ IO ½Ã½ºÅÛ ÄÝ¿¡ µû¸¥ ¿î¿µÃ¼Á¦ ¹®¸ÆÀüȯ ¿À¹öÇìµå¸¦ À¯¹ßÇÏ°Ô µÈ´Ù. ±âÁ¸ À妽º ½ºÄµ ¹æ½ÄÀÇ ÀÌ·¯ÇÑ ´ÜÁ¡À» ±Øº¹Çϱâ À§ÇØ, º» ³í¹®¿¡¼­´Â À妽º ½ºÄµ ½Ã¿¡ ºñµ¿±â ´ÙÁß ÆäÀÌÁö Àб⠱â¹ýÀ» Á¦¾ÈÇÑ´Ù. Á¦¾È ±â¹ýÀ» PostgreSQL ¿ÀǼҽº DBMS¸¦ ÀÌ¿ëÇؼ­ ±¸ÇöÇÏ¿´°í, TPC-E º¥Ä¡¸¶Å©¸¦ ÀÌ¿ëÇÑ ½ÇÇè °á°ú Á¦¾È ±â¹ýÀº ±âÁ¸ µ¿±âÀû À妽º ½ºÄµ¿¡ ºñÇØ 50% ÀÌ»óÀÇ ¼º´É Çâ»óÀ» º¸¿´´Ù.
¿µ¹®³»¿ë
(English Abstract)
In traditional index scan methods adopted by all modern DBMSs, data pages pointed by index entries in index leaf nodes are read into database buffer one by one in a synchronous way. Therefore, in case of non-clustered indexes where data records satisfying the given search condition are scattered over multiple data pages, the existing index scan mechanism cannot exploit the parallelism inside modern flash SSD and incurs excessive OS context switch overhead due to numerous IO system calls. In order to overcome these limitations of the existing index scan mechanism, this paper proposes a new index scan supporting asynchronous multiple page reads. We implemented the proposed scheme using PostgreSQL, an open source DBMS, and our experiment using the TPC-E benchmark shows that our scheme can improve the index scan performance over the existing synchronous index scan scheme by more than fifty percents.
Å°¿öµå(Keyword) B+-tree À妽º ½ºÄµ   ºñµ¿±â IO   Ç÷¡½Ã SSD   PostgreSQL   B+-tree index scan   asynchronous IO   flash SSD   PostgreSQL  
ÆÄÀÏ÷ºÎ PDF ´Ù¿î·Îµå