Á¤º¸°úÇÐȸ ³í¹®Áö A : ½Ã½ºÅÛ ¹× ÀÌ·Ð
Current Result Document :
ÇѱÛÁ¦¸ñ(Korean Title) |
ºÐ»ê ½Ã½ºÅÛ¿¡¼ÀÇ À¯ÀüÀÚ ½ºÄÉÁ층 ¾Ë°í¸®Áò |
¿µ¹®Á¦¸ñ(English Title) |
Genetic Scheduling Algorithms in Distributed Computing Systems |
ÀúÀÚ(Author) |
¿ì¼ºÈ£
¾ç¼ººÀ
±è½Å´ö
ÇÑŹµ·
Sungho Woo
Sungbong Yang
Shindug Kim
Tackdon Han
|
¿ø¹®¼ö·Ïó(Citation) |
VOL 24 NO. 12 PP. 1247 ~ 1256 (1997. 12) |
Çѱ۳»¿ë (Korean Abstract) |
ºÐ»ê ½Ã½ºÅÛ¿¡¼ ÀÛ¾÷µé°£ÀÇ ¿ì¼± ¼øÀ§(precedence relation)¸¦ À§ÇØ ¹æÇ⼺ ºñ¼øȯ ±×·¡ÇÁ(Directed Acyclic Graph)·Î Ç¥ÇöµÈ º´·Ä ÇÁ·Î±×·¥À» ½ºÄÉÁìÇÏ´Â ¹®Á¦´Â Á¦ÇÑµÈ °æ¿ì¸¦ Á¦¿ÜÇÏ°í´Â NP-complete ¹®Á¦·Î ¾Ë·ÁÁ® ÀÖ´Ù[1]. µû¶ó¼ ½ÇÁ¦ ½Ã½ºÅÛ¿¡¼ÀÇ ÀÛ¾÷ ½ºÄÉÁ층À» À§ÇØ ´Ù¾çÇÑ ¸ðµ¨°ú °¡Á¤ ÇÏ¿¡¼ ÈÞ¸®½ºÆ½¿¡ ±â¹ÝÀ» µÐ ¾Ë°í¸®ÁòµéÀÌ Á¦¾ÈµÇ¾î ¿Ô´Ù[2,3,4,5,6,7,8]. ºÐ»ê ½Ã½ºÅÛÀº ÇÁ·Î¼¼¼µé°ú ³×Æ®¿öÅ©ÀÇ ¼ºÁú¿¡ µû¶ó Å©°Ô µ¿±âÁ¾ ºÐ»ê ½Ã½ºÅÛ(Distributed Homogeneous System)°ú À̱âÁ¾ ºÐ»ê ½Ã½ºÅÛ(Distributed Heterogeneous System)À¸·Î ±¸ºÐ µÉ ¼ö ÀÖ´Ù. º» ³í¹®¿¡¼´Â ºÐ»ê ½Ã½ºÅÛÀÇ ÀϹÝÀûÀÎ ¸ðµ¨À» µ¿±âÁ¾ ºÐ»ê ½Ã½ºÅÛ°ú À̱âÁ¾ ºÐ»ê ½Ã½ºÅÛ¿¡ ´ëÇØ Á¤ÀÇÇÏ°í, Á¤ÀÇµÈ ºÐ»ê ½Ã½ºÅ۵鿡¼ ÀÛ¾÷ ½ºÄÉÁÙ¸µÀ» ¼öÇàÇÏ´Â À¯ÀüÀÚ ½ºÄÉÁ층 ¾Ë°í¸®Áò(Genetic Scheduling Algorithms, GSA)À» Á¦¾ÈÇÑ´Ù. µ¿±âÁ¾ ºÐ»ê ½Ã½ºÅÛ°ú À̱âÁ¾ ºÐ»ê ½Ã½ºÅÛ¿¡¼ÀÇ ÀÛ¾÷ ½ºÄÉÁ층Àº °¢ ½Ã½ºÅÛÀÇ »óÀÌÇÑ ¼ºÁú ¶§¹®¿¡ ´Ù¸¥ ¹®Á¦·Î °£Áֵǰí ÀÖ´Ù. ±×·¯³ª GSA´Â À¯ÀüÀÚ ¾Ë°í¸®ÁòÀÌ Áö´Ñ °ß°í(robust)ÇÑ ¼ºÁúÀ» ÅëÇØ ºÐ»ê ½Ã½ºÅÛÀÇ º¯È¿¡ µû¶ó ½ºÄÉÁ층ÀÇ ¼öÇà½Ã¿¡ °í·ÁÇØ¾ß ÇÒ ¿äÀÎÀÌ º¯ÇÏ´Â ¿µÇâÀ» ±Øº¹ÇÒ ¼ö ÀÖ´Ù. º» ³í¹®¿¡¼ Á¦¾ÈµÈ GSA´Â ¸ðÀÇ ½ÇÇèÀ» ÅëÇؼ µ¿±âÁ¾ ½Ã½ºÅÛ¿¡¼´Â ¸®½ºÆ® ½ºÄÉÁ층 ¾Ë°í¸®Áò[5]°ú, À̱âÁ¾ ºÐ»ê ½Ã½ºÅÛ¿¡¼´Â OLROG(One-Level Reach-Out Greedy) ¾Ë°í¸®Áò[7]°ú ¼º´ÉÀ» ºñ±³ ¹× ºÐ¼®ÇÏ¿´´Ù. GSA´Â ´Ù¾çÇÑ È¯°æ¿¡¼ ¶Ù¾î³ ¼º´ÉÀÇ Çâ»óÀ» º¸¿© ÁÖ¾ú´Ù.
|
¿µ¹®³»¿ë (English Abstract) |
Scheduling a directed acyclic graph (DAG) which represents the precedence relations of the tasks of a parallel program in a distributed computing system (DCS) is known as an NP-complete problem except for some special cases[1]. Many heuristic-based methods have been proposed under various models and assumptions[2,3,4,5,6,7,8].
A DCS can be classified into two types according to the characteristics of the processors on a network: a distributed homogeneous system(DHOS) and a distributed heterogeneous system(DHES). This paper defines a general model for a DHOS and a DHES and presents genetic scheduling algorithms(GSA) to solve the task scheduling problem in the defined distributed computing systems. Task scheduling problems in DHOS and DHES are regarded as different problems due to their different factors in the scheduling. But the innate robustness of GSA can overcome the different factors. The performances of our GSA are compared with the list scheduling algorithm [3,5] in a DHOS and with the one-level reach-out greedy algorithm (OLROG) [7] in a DHES. The GSA presented in this paper have shown better performance in most cases than other scheduling methods.
|
Å°¿öµå(Keyword) |
|
ÆÄÀÏ÷ºÎ |
PDF ´Ù¿î·Îµå
|