µ¥ÀÌÅͺ£À̽º ¿¬±¸È¸Áö(SIGDB)
ÇѱÛÁ¦¸ñ(Korean Title) |
Spark ȯ°æ¿¡¼ ½ºÆ®¸² µ¥ÀÌÅÍ Ã³¸®¸¦ À§ÇÑ È¿À²ÀûÀÎ ½ºÄÉÁÙ¸µ ±â¹ý |
¿µ¹®Á¦¸ñ(English Title) |
An Efficient Scheduling Scheme for Data Stream Processing in Spark Environments |
ÀúÀÚ(Author) |
ÀüÇö¿í
±è¹Î¼ö
¼ÛÁø¿ì
ÃÖµµÁø
±è¿¬¿ì
ÀÓÁ¾ÅÂ
º¹°æ¼ö
À¯Àç¼ö
Hyeonwook Jeon
Minsoo Kim
JinWoo Song
DoJin Choi
Yeonwoo Kim
Jongtae Lim
Kyoungsoo Bok
Jaesoo Yoo
|
¿ø¹®¼ö·Ïó(Citation) |
VOL 32 NO. 02 PP. 0076 ~ 0088 (2016. 08) |
Çѱ۳»¿ë (Korean Abstract) |
ÃÖ±Ù IT ±â¼úÀÇ ¹ß´Þ°ú ÇÔ²² ¼Ò¼È ¹Ìµð¾î, ¸ð¹ÙÀÏ ´Ü¸»±â, »ç¹°ÀÎÅͳݰú °°Àº ´Ù¾çÇÑ ¸Åü·Î ÀÎÇØ ´ë±Ô¸ð·Î ¹ß»ýÇÏ´Â ½ºÆ®¸®¹Ö ºòµ¥ÀÌÅ͸¦ ½Ç½Ã°£ ó¸®Çϱâ À§ÇÑ ¸¹Àº ¿¬±¸µéÀÌ ÁøÇàµÇ°í ÀÖ´Ù. ½ºÆ®¸² µ¥ÀÌÅ͸¦ ½Ç½Ã°£ ó¸®Çϱâ À§Çؼ´Â ºÐ»ê Àâ ½ºÄÉÁÙ¸µ ±â¹ýÀÌ ¸Å¿ì Áß¿äÇÏ´Ù. º» ³í¹®¿¡¼´Â Spark¿¡¼ ½ºÆ®¸² µ¥ÀÌÅ͸¦ ½Ç½Ã°£ ó¸®Çϱâ À§ÇØ ³ëµåÀÇ ºÎÇϸ¦ °í·ÁÇÑ È¿À²ÀûÀÎ ½ºÄÉÁÙ¸µ ±â¹ýÀ» Á¦¾ÈÇÑ´Ù. Á¦¾ÈÇÏ´Â ±â¹ý¿¡¼´Â °¢ ³ëµåÀÇ ºÎÇϸ¦ ÆÇ´ÜÇϱâ À§ÇØ CPU »ç¿ë·®, ¸Þ¸ð¸® ºÎÇÏ, Æò±ÕÀÀ´ä½Ã°£À» °í·ÁÇÑ´Ù. ³ëµåÀÇ ºÎÇÏ¿¡ µû¶ó ÀÛ¾÷À» ÇÒ´çÇÏ°í ÇÒ´çµÈ ÀÛ¾÷ÀÇ º¹Àâµµ·Î ÀÎÇØ ³ëµåÀÇ ºÎÇÏ°¡ Áõ°¡µÉ °æ¿ì ºÎÇÏ°¡ ÀûÀº ³ëµå¿¡ ÀÛ¾÷À» º¹Á¦ÇÏ¿© ÀÛ¾÷ 󸮸¦ ¼öÇàÇÔÀ¸·Î½á Áö¿¬À» ¹æÁöÇÒ ¼ö ÀÖ´Ù. Á¦¾ÈÇÏ´Â ±â¹ýÀÇ ¿ì¼ö¼ºÀ» Áõ¸íÇϱâ À§ÇØ ±âÁ¸ ±â¹ý°úÀÇ ´Ù¾çÇÑ ¼º´ÉÆò°¡¸¦ ¼öÇàÇÑ´Ù. |
¿µ¹®³»¿ë (English Abstract) |
Recently, studies on the real-time processing of big data stream generated through various media such as social media, mobile device, and internet of things along with the development of IT technologies have been done. In order to process the data streams in real-time, a distributed job scheduling scheme is very important. In this paper, we propose an efficient scheduling scheme considering node loads for the real-time data stream processing in Spark environments. The proposed scheme considers CPU utilization, memory loads, and average response times in order to determine the load of each node. It can protect the processing delay by assigning jobs to nodes according to their loads and by replicating the jobs to the nodes with little loads when the node loads increase due to the complexity of the assigned jobs. In order to show the superiority of the proposed scheme, we compare it with the existing schemes through various performance evaluations. |
Å°¿öµå(Keyword) |
½ºÆÄÅ©
½ºÄÉÁÙ¸µ
½ºÆ®¸² µ¥ÀÌÅÍ
»ç¹°ÀÎÅͳÝ
Spark
scheduling
data stream
IoT
|
ÆÄÀÏ÷ºÎ |
PDF ´Ù¿î·Îµå
|