• 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 :

ÇѱÛÁ¦¸ñ(Korean Title) LVS-Cache: Low-Degree Vertex Static Cache¸¦ ÅëÇÑ µð½ºÅ© ±â¹Ý ±×·¡ÇÁ ¿£Áø ¼º´É Çâ»ó
¿µ¹®Á¦¸ñ(English Title) LVS-Cache: Performance Improvement using Low-Degree Vertex Static Caching on Disk-based Graph Engine
ÀúÀÚ(Author) ±èÁ¤Çö   ÀÌÀºÀç   ÀÓ±ÙÇР  ¼­Áö¿ø   ³ë»ïÇõ   Junghyun Kim   Eunjae Lee   Keunhak Lim   Jiwon Seo   Sam H. Noh   ±èÁ¤Çö   ÀÌÀºÀç   ÀÓ±ÙÇР  ¼­Áö¿ø   ³ë»ïÇõ   Junghyun Kim   Eunjae Lee   Keunhak Lim   Jiwon Seo   Sam H. Noh  
¿ø¹®¼ö·Ïó(Citation) VOL 24 NO. 11 PP. 0628 ~ 0633 (2018. 11)
Çѱ۳»¿ë
(Korean Abstract)
À¥°ú ¼Ò¼È ³×Æ®¿öÅ©ÀÇ ¹ßÀü¿¡ µû¶ó ´ë±Ô¸ð ±×·¡ÇÁ¸¦ È¿°úÀûÀ¸·Î ó¸®ÇÏ´Â ±â¹ý¿¡ ´ëÇÑ ¿¬±¸°¡ È°¹ßÈ÷ ÁøÇàµÇ°í ÀÖ´Ù. ±× Áß, ´ÜÀÏ ¸Ó½Å¿¡¼­ ÀÛÀº ¸Þ¸ð¸®·Îµµ ±×·¡ÇÁ 󸮰¡ °¡´ÉÇÑ, µð½ºÅ© ±â¹Ý ±×·¡ÇÁ ¿£ÁøÀÌ ¸¹ÀÌ È°¿ëµÇ°í ÀÖ´Ù. ÇÏÁö¸¸ ÀÌ ¿£ÁøÀº µð½ºÅ© ÀÔÃâ·Â ¼º´ÉÀ¸·Î ÀÎÇÑ Àüü ÇÁ·Î¼¼½Ì ¼º´É ÀúÇÏ°¡ ÀÖ´Ù. ±âÁ¸ÀÇ ½Ã½ºÅÛÀº ºí·Ï ij½Ã¸¦ È°¿ëÇÏ¿© ÀÌ·¯ÇÑ ¹®Á¦¸¦ ÇØ°áÇÏ°íÀÚ ÇÏ¿´À¸³ª, ±×·¡ÇÁ ó¸®ÀÇ Æ¯¼º»ó, ºí·Ï ij½Ã´Â ij½Ã ÀûÁß·üÀÌ ³·¾Æ È¿À²¼ºÀÌ ¶³¾îÁö°í, µ¿ÀûÀ¸·Î ÇÒ´çÇÏ°í ÇØÁ¦ÇÏ´Â ºñ¿ëÀÌ ¹ß»ýÇÑ´Ù. º» ³í¹®¿¡¼­´Â ÇØ´ç ¹®Á¦ÀÇ ÇØ°áÃ¥À¸·Î LVS-Cache ±â¹ýÀ» Á¦¾ÈÇÑ´Ù. ÀÌ ±â¹ýÀº ºí·Ï ij½ÃÀÇ È¿À²¼ºÀ» ÀúÇϽÃÅ°´Â µ¥ÀÌÅ͸¦ ¹Ì¸® Á¤ÀûÀ¸·Î ij½ÌÇÑ´Ù. Degree°¡ ³·Àº VertexµéÀ» ij½ÌÇÔÀ¸·Î½á, ¸Þ¸ð¸® »ç¿ë·® ´ëºñ ³ôÀº ij½Ã È°¿ë È¿°ú¸¦ ¾ò°íÀÚ ÇÏ¿´´Ù. Á¦¾ÈÇÑ LVS-Cache ¿¡ ´ëÇÑ ½ÇÇè °á°ú, ºí·Ï ij½Ã¿¡ ºñÇØ ¼º´ÉÀÌ Çâ»óµÇ¾úÀ¸¸ç, µð½ºÅ© ÀÔÃâ·Â ¾çµµ È¿°úÀûÀ¸·Î ÁÙ¾îµç °ÍÀ» È®ÀÎÇÏ¿´´Ù.
¿µ¹®³»¿ë
(English Abstract)
Advancements in the web and social network have led to the flooding of large-scale graph data. Recently, researches to efficiently deal with such a big-data were proposed. Among them, disk-based graph engines were designed to process large-scale graphs on a single machine. These engines can process large-scale graph using small memory consumption. However, the engines should read graph data from disk, so disk I/O performance is critical for the whole graph processing performance. The state-of-the-art systems have proposed block-cache to reduce the direct I/O to disk. However, because of the characteristics of graph processing, the hit ratio of block-cache is low, which means that the utilization of block-cache is not good. Moreover, because it's also "cache", it needs dynamic allocation and eviction cost. In this paper, we propose LVS-Cache (Low-Degree Vertex Static Cache) to optimize disk I/O performance. LVS-Cache allocates vertices statically that degrades block-cache's utilization. Among the vertices that need to be allocated to LVS-Cache, we have chosen low-degree vertices as prior ones which have higher cache utilization in proportion to memory consumption. The result reveals improvement in performance compared to block-cache and reduction in the amount of I/O request to disk.
Å°¿öµå(Keyword) ±×·¡ÇÁ ¿£Áø   ±×·¡ÇÁ 󸮠  ´ë±Ô¸ð ±×·¡ÇÁ   ij½Ì Á¤Ã¥   graph engine   graph processing   large-scale graph   caching policy   ±×·¡ÇÁ ¿£Áø   ±×·¡ÇÁ 󸮠  ´ë±Ô¸ð ±×·¡ÇÁ   ij½Ì Á¤Ã¥   graph engine   graph processing   large-scale graph   caching policy  
ÆÄÀÏ÷ºÎ PDF ´Ù¿î·Îµå