Á¤º¸°úÇÐȸ ÄÄÇ»ÆÃÀÇ ½ÇÁ¦ ³í¹®Áö (KIISE Transactions on Computing Practices)
ÇѱÛÁ¦¸ñ(Korean Title) |
Ŭ·¯½ºÅÍ ±â¹Ý °úÇÐÀû °¡½ÃÈ µµ±¸¸¦ À§ÇÑ °¡½ÃÈ ÆÄÀÌÇÁ¶óÀÎ ¼³°è ¹× ±¸Çö |
¿µ¹®Á¦¸ñ(English Title) |
Visualization Pipeline Design and Implementation for Cluster-based Scientific Visualization Tools |
ÀúÀÚ(Author) |
ÀÌÁß¿¬
±è´ö¼ö
Joong-Youn Lee
Duksu Kim
|
¿ø¹®¼ö·Ïó(Citation) |
VOL 25 NO. 06 PP. 0285 ~ 0293 (2019. 06) |
Çѱ۳»¿ë (Korean Abstract) |
º» ³í¹®Àº Ŭ·¯½ºÅÍ ±â¹Ý °úÇÐÀû °¡½ÃÈ µµ±¸¸¦ À§ÇÑ °¡½ÃÈ ÆÄÀÌÇÁ¶óÀÎÀÇ ¼³°è¸¦ Á¦¾ÈÇÏ°í ±× ±¸Çö »ç·Ê¸¦ °øÀ¯ÇÑ´Ù. º» ¿¬±¸ÀÇ ÆÄÀÌÇÁ¶óÀÎÀº °¢°¢ÀÇ °¡½ÃÈ ±â¹ýÀ» Ç¥ÇöÇÏ´Â °¡½ÃÈ ÆÄÀÌÇÁµé·Î ±¸¼ºµÇ¸ç, °¡½ÃÈ ÆÄÀÌÇÁµéÀº Á¤ÇüÈµÈ ÆÄÀÌÇÁ ±¸Á¶¸¦ »ç¿ëÇÔÀ¸·Î¼ »óÈ£ À¯¿¬ÇÑ ¿¬°áÀ» º¸ÀåÇÑ´Ù. ¶ÇÇÑ ÆÄÀÌÇÁ¶óÀÎÀ» »ç¿ëÇÏ´Â °¡½ÃÈ µµ±¸¿Í Åë½ÅÀ» Àü´ãÇϴ Ư¼ö ÆÄÀÌÇÁ(ÀÎÅÍÆäÀ̽º ÆÄÀÌÇÁ)¸¦ »ç¿ëÇÔÀ¸·Î¼, °¡½ÃÈ µµ±¸¿¡ µ¶¸³ÀûÀÎ ±¸Çö ¹× °ü¸®°¡ °¡´ÉÇÏ´Ù. Á¦¾ÈµÈ °¡½ÃÈ ÆÄÀÌÇÁ¶óÀÎÀº ÀÎÅÍÆäÀ̽º ÆÄÀÌÇÁµéÀ» Á¦¿ÜÇÑ ÇÏÀ§ ÆÄÀÌÇÁ¶óÀÎÀ» °¢ ¿¬»ê ³ëµå¿¡ ¹èÄ¡ÇÏ°í, À̵éÀ» ¸¶½ºÅÍ ³ëµå¿¡ ÀÖ´Â ÀÎÅÍÆäÀ̽º ÆÄÀÌÇÁ¿¡ ¿¬°áÇÔ À¸·Î¼ Ŭ·¯½ºÅÍ ±â¹Ý °¡½ÃÈ ÆÄÀÌÇÁ¶óÀÎÀ¸·Î °£´ÜÈ÷ È®ÀåµÈ´Ù. ÆÄÀÌÇÁ¶óÀÎÀº »ç¿ëÀÚÀÇ °¡½ÃÈ ¿äûÀ» ó¸®ÇÒ ¼ö ÀÖ´Â ÆÄÀÌÇÁÀÇ Á¶ÇÕÀ¸·Î ½Ç½Ã°£À¸·Î »ý¼ºµÈ´Ù. º» ¿¬±¸´Â Á¦¾ÈÇÏ´Â °¡½ÃÈ ÆÄÀÌÇÁ¶óÀÎÀ» Ŭ·¯½ºÅÍ ±â¹Ý °¡½ÃÈ ÇÁ·¹ÀÓ¿öÅ© Áß ÇϳªÀÎ GLOVE¸¦ ±â¹ÝÀ¸·Î 32°³ÀÇ ¿¬»ê³ëµå·Î ±¸¼ºµÈ Ŭ·¯½ºÅÍ¿¡ ±¸ÇöÇÏ¿´À¸¸ç, ³× Á¾·ùÀÇ ´ë¿ë·® °úÇе¥ÀÌÅÍÀÇ °¡½ÃÈ¿¡ Àû¿ëÇÏ¿´´Ù. ±× °á°ú ´Ù¾çÇÑ Á¶ÇÕÀÇ °¡½ÃÈ ¿äûÀ» ¼º°øÀûÀ¸·Î ó¸®ÇÏ¿´À¸¸ç, ¿¬»ê ³ëµå ¼ö Áõ°¡¿¡µµ ¾ÈÁ¤ÀûÀ¸·Î µ¿ÀÛÇÏ´Â ¸ð½ÀÀ» È®ÀÎ ÇÒ ¼ö ÀÖ¾ú´Ù. ¶ÇÇÑ µ¿ÀÏÇÑ Á¶ÇÕÀÇ °¡½ÃÈ ¿äû¿¡ ´ëÇØ, ÆÄÀÌÇÁ¶óÀÎÀ» Àû¿ëÇϱâ Àü GLOVE ´ëºñ ÃÖ´ë 7.72¹è ³ôÀº °¡½ÃÈ ¼º´ÉÀ» º¸¿©ÁÖ¾ú´Ù. ÀÌ´Â Á¦¾ÈÇÑ °¡½ÃÈ ÆÄÀÌÇÁ¶óÀÎÀÇ È¿¿ë¼º ¹× È®À强À» º¸¿©ÁÖ´Â °á°ú´Ù.
|
¿µ¹®³»¿ë (English Abstract) |
We propose a visualization pipeline design for cluster-based scientific visualization tools and share an implementation example. Our visualization pipeline is composed of visualization pipes that represent each visualization method. Visualization pipes follow a unified structure, and this structure guarantees flexible connectivity among pipes. By using dedicated pipes (i.e., interface pipes) for communication between our pipeline and visualization tools using that, we can implement visualization pipes independently with the tools. The proposed visualization pipeline is simply extended to a cluster-based visualization pipeline, by building a sub-pipeline consisting of only visualization pipes on each slave node and connecting with interface pipes in the master node. A visualization pipeline is built on the fly according to the visualization request from users. We applied our visualization pipeline design to an existing cluster-based visualization framework (GLOVE) and implemented it in a cluster with 32 computing nodes (i.e., slave nodes). Then we applied it to four different large scientific datasets, and found that our cluster-based visualization pipeline successfully handles various combinations of visualization queries. Also, it remained stable, as the number of slave nodes increased. For a given combination of visualization queries, GLOVE with our pipeline shows up to 7.72 times higher performance, compared with the original GLOVE. These results demonstrate the usefulness and scalability of our visualization pipeline design.
|
Å°¿öµå(Keyword) |
°úÇÐÀû °¡½ÃÈ
°¡½ÃÈ ÆÄÀÌÇÁ¶óÀÎ
ÄÄÇ»Æà Ŭ·¯½ºÅÍ
°úÇÐ µ¥ÀÌÅÍ
ºÐ»êó¸®
scientific visualization
visualization pipeline
computing cluster
scientific data
distributed computing
|
ÆÄÀÏ÷ºÎ |
PDF ´Ù¿î·Îµå
|