Á¤º¸°úÇÐȸ ÄÄÇ»ÆÃÀÇ ½ÇÁ¦ ³í¹®Áö (KIISE Transactions on Computing Practices)
Current Result Document : 164 / 164
ÇѱÛÁ¦¸ñ(Korean Title) |
ºòµ¥ÀÌÅÍ ºÐ¼® ¹× ¸Ó½Å·¯´× ¼ºñ½º¸¦ À§ÇÑ ÄÁÅ×ÀÌ³Ê ±â¹Ý HPC Ŭ·¯½ºÅÍ ±¸Ãà ¹× ¼º´É ºÐ¼® |
¿µ¹®Á¦¸ñ(English Title) |
The Implementation and Performance Analysis of Container-based HPC Cluster System for Big Data Analysis and Machine Learning Services |
ÀúÀÚ(Author) |
¹ÚÁÖ¿ø
À̽¹Î
È«Å¿µ
Ju-Won Park
Seungmin Lee
Taeyoung Hong
|
¿ø¹®¼ö·Ïó(Citation) |
VOL 27 NO. 05 PP. 0228 ~ 0233 (2021. 05) |
Çѱ۳»¿ë (Korean Abstract) |
±â»ó, ÈÇÐ, °í¿¡³ÊÁö ¹°¸®¿Í °°Àº ±âÃÊ °úÇÐ ºÐ¾ßÀÇ ´ë±Ô¸ð ¿öÅ©·Îµå¸¦ ½ÇÇàÇϱâ À§ÇØ ´ë±Ô¸ðÀÇ Å¬·¯½ºÅÍ ½Ã½ºÅÛ¿¡¼´Â º´·Ä ÄÄÇ»Æÿ¡ ÃÖÀûÈµÈ ½Ã½ºÅÛ È¯°æÀ» ±¸ÃàÇÏ¿© ¼ºñ½º¸¦ Á¦°øÇÏ°í ÀÖ´Ù. ±×·¯³ª ÃÖ±Ù 4Â÷ »ê¾÷ Çõ¸íÀ¸·Î ÀÎÇÏ¿© ºòµ¥ÀÌÅÍ ºÐ¼®, ¸Ó½Å ·¯´×°ú °°Àº »õ·Î¿î ÇüÅÂÀÇ »ç¿ëÀÚ ¿ä±¸ »çÇ×ÀÌ Á¦½ÃµÊ¿¡ µû¶ó ±âÁ¸ Ŭ·¯½ºÅÍ ½Ã½ºÅÛ¿¡¼´Â °æÁ÷µÈ ¼ÒÇÁÆ®¿þ¾î ±¸¼º, À¯Áö º¸¼ö ºñ¿ë Áõ°¡, Á¦ÇÑµÈ ÀÛ¾÷ ÇüÅ Áö¿ø µî ¸¹Àº ¾î·Á¿ò¿¡ Á÷¸éÇØ ÀÖ´Ù. ÀÌ·¯ÇÑ ¾î·Á¿òÀ» ÇØ°áÇϱâ À§ÇØ º» ³í¹®¿¡¼´Â ÄÁÅ×ÀÌ³Ê ±â¹Ý HPC Ŭ·¯½ºÅÍ ±¸Ãà ¹æ¾ÈÀ» Á¦½ÃÇÑ´Ù. Á¦½ÃµÈ ±¸Ãà ¹æ¾È¿¡¼´Â ÄÁÅ×ÀÌ³Ê ±â¼úÀ» È°¿ëÇÏ¿© Ŭ·¯½ºÅÍ ½Ã½ºÅÛÀÇ ¼ÒÇÁÆ®¿þ¾î ½ºÅÃÀ» °£°áÈÇÔÀ¸·Î½á ÀÚ¿ø À¯¿¬¼ºÀ» Çâ»ó½ÃÄÑ »ç¿ëÀÚ ¸ÂÃãÇü ¼ºñ½º°¡ °¡´ÉÇÏ´Ù. ½ÇÇè ȯ°æÀº ÄÁÅ×ÀÌ³Ê ½ÇÇà µ¥¸óÀ¸·Î Docker, ÄÁÅ×ÀÌ³Ê °ü¸® ÅøŶÀ¸·Î´Â kubernetes¸¦ È°¿ëÇÏ¿© ±¸ÃàÇßÀ¸¸ç HPL, HPCG, NPB benchmark ÅøŶÀ» ÀÌ¿ëÇÏ¿© ´Ù¾çÇÑ ¿öÅ©·Îµå¿¡ ´ëÇÑ °è»ê ¼º´É ÃøÁ¤ °á°ú¸¦ Á¦½ÃÇÑ´Ù.
|
¿µ¹®³»¿ë (English Abstract) |
Several fields of basic sciences, such as meteorology, chemistry, and high-energy physics, require large-scale workload. To support it, a high-performance computing (HPC) system, which is optimized for parallel computing, is widely used. Recently, due to the fourth industrial revolution, there has been an increasing demand for HPC resources in computer science fields, such as big data analysis and machine learning. To support these new services, the existing HPC system has major problems, such as rigid software configuration, high maintenance cost, and limited job type support. In this paper, we propose a container-based HPC cluster system architecture to solve these problems. The presented system architecture uses Docker for container runtime engine and kubernetes as the container orchestration toolkit.
|
Å°¿öµå(Keyword) |
ÄÁÅ×À̳Ê
kubernetes
HPC Ŭ·¯½ºÅÍ ½Ã½ºÅÛ
º¥Ä¡¸¶Å© ÅøŶ
container
kubernetes
HPC cluster system
benchmark
|
ÆÄÀÏ÷ºÎ |
PDF ´Ù¿î·Îµå
|