• 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) SCMÀ» Àû¿ëÇÑ ºòµ¥ÀÌÅÍ ½ºÆ®¸² ó¸® ½Ã½ºÅÛ ¼³°è ¹× ¼º´É ºÐ¼®
¿µ¹®Á¦¸ñ(English Title) Bigdata Streaming Eco System Design and Performance Analysis using SCM
ÀúÀÚ(Author) ÀÌ°æ¼ö   ±è¼ÒÇö   ¼Û´ÙÀº   Kyungsoo Lee   Sohyun Kim   Daeun Song  
¿ø¹®¼ö·Ïó(Citation) VOL 25 NO. 11 PP. 0523 ~ 0532 (2019. 11)
Çѱ۳»¿ë
(Korean Abstract)
DRAMÀÇ °íºñ¿ë°ú ½ºÅ丮ÁöÀÇ IO ¿À¹öÇìµå°¡ ÀÎ-¸Þ¸ð¸® ±â¹ÝÀÇ ºòµ¥ÀÌÅÍ ½ºÆ®¸² ó¸® ½Ã½ºÅÛ ±¸Ãà¿¡ ¹®Á¦°¡ µÇ°í ÀÖ´Ù. º» ³í¹®Àº ½ºÆ®¸² ó¸® ½Ã½ºÅÛÀÇ ºñ¿ë, ¼º´É ¹®Á¦¸¦ ÇØ°áÇϱâ À§ÇØ SCMÀ» °í¿ë·®ÀÇ ¸Þ¸ð¸®¿Í ºü¸¥ ½ºÅ丮Áö·Î Àû¿ëÇÏ´Â ¹æ¾ÈÀ» ¼Ò°³ÇÑ´Ù. ¸Þ¸ð¸® ¸ðµå´Â DRAMÀ» SCMÀ¸·Î ´ëüÇØ ºñ¿ëÀ» ³·Ãß°í, ½ºÅ丮Áö ¸ðµå´Â Disk ´ë½Å SCMÀ» »ç¿ëÇØ ½Ã½ºÅÛ ¼º´ÉÀ» ³ôÀδÙ. º» ³í¹®Àº Spark Streaming ECO SystemÀ» ±¸¼ºÇÏ´Â ÀÀ¿ë º° SCMÀ» Àû¿ëÇØ DRAM Àý°¨°ú ¼º´É º¯È­¸¦ ÃøÁ¤ÇÏ°í, ÀÌ ÀÀ¿ëÀÌ Àüü ½Ã½ºÅÛ¿¡ ¹ÌÄ¡´Â ¿µÇâÀ» ºÐ¼®ÇÏ¿´´Ù. ¸ðµç ÀÀ¿ë¿¡ ¸Þ¸ð¸® ¸ðµå, ½ºÅ丮Áö ¸ðµå¸¦ µ¿½Ã Àû¿ëÇÑ °á°ú QoS´Â 95th´Â 1.4%, 99th´Â 2.8% Áõ°¡ÇÏ¿´°í, DRAMÀº 31% Àý°¨Çß´Ù. Spark ¸Þ¸ð¸® ¸ðµå°¡ ½Ã½ºÅÛ¿¡ ¹ÌÄ¡´Â ¿µÇâÀÌ °¡Àå Å©¸ç, SCMÀÌ Àû¿ëµÈ Ÿ ÀÀ¿ëÀÇ ¼º´É º¯È­´Â SparkÀ¸·Î ÀÎÇØ µå·¯³ªÁö ¾Ê´Â °ÍÀ¸·Î ºÐ¼®µÇ¾ú´Ù.
¿µ¹®³»¿ë
(English Abstract)
The high cost of the DRAM and the IO overhead of the storage has become a problem of In-memory based stream system construction. This paper introduces SCM as a high capacity memory and fast storage to solve cost and performance problems of the stream processing system. The memory mode of the SCM replaces the heap of the DRAM with the SCM to lower the cost of the DRAM, and the storage mode of the SCM uses SCM instead of Disk to improve system performance. The big data stream processing system to apply SCM selected Spark Streaming ECO System. In the present work, we have analyzed the effect of SCM application on the whole system by measuring performance change and the reduction of DRAM by applying SCM for each application constituting the system. Application of both memory and storage modes to all applications resulted in a 1.4%/2.8% increase in QoS performance at 95th/99th, and a 31% reduction in DRAM. In addition, it was analyzed that the memory mode applied to Spark has the greatest influence on the system. The change in performance of other applications using SCM was not revealed due to Spark.
Å°¿öµå(Keyword) ºòµ¥ÀÌÅÍ   ½ºÆÄÅ©   ½ºÆ®¸² ó¸® ½Ã½ºÅÛ   ºòµ¥ÀÌÅÍ ¿¡ÄÚ ½Ã½ºÅÛ   ½ºÅ丮Áö Ŭ·¡½º ¸Þ¸ð¸®   big data   spark   stream processing   big data eco system   storage class memory  
ÆÄÀÏ÷ºÎ PDF ´Ù¿î·Îµå