ÇѱÛÁ¦¸ñ(Korean Title) |
½Ã½ºÅÛ Àü¿ª ·Î±× µ¥ÀÌÅ͸¦ »ç¿ëÇÑ °í¼º´É ÄÄÇ»Æà ½Ã½ºÅÛ ÀÀ¿ë I/O Ư¼º ºÐ¼® |
¿µ¹®Á¦¸ñ(English Title) |
Analysis of Application I/O Characteristics in HPC Systems with System-wide Log Data |
ÀúÀÚ(Author) |
¼ºµ¿±Ô
¹æÁö¿ì
±èÁ¤¿ë
±è¼º°ï
¾öÇö»ó
È«Å¿µ
Dong Kyu Sung
Jiwoo Bang
Chungyong Kim
Sunggon Kim
Hyeonsang Eom
TaeYoung Hong
|
¿ø¹®¼ö·Ïó(Citation) |
VOL 48 NO. 02 PP. 0019 ~ 0021 (2021. 12) |
Çѱ۳»¿ë (Korean Abstract) |
|
¿µ¹®³»¿ë (English Abstract) |
High performance computing (HPC) systems consist of thousands of compute, storage, and network nodes which have capabilities of handling simultaneous access by a large number of users. In such systems, it is difficult to analyze I/O behaviors and performance of HPC applications due to diversity of the applications and multiple software stacks that the applications go through during executions. In order to understand I/O characteristics of the applications, it is important to analyze I/O-related data that covers all layers and modules that participate in executions of the applications. This paper presents an integrated database generation from different I/O monitoring tools and analysis of I/O characteristics of real-world applications executed in KISTI Nurion supercomputer. |
Å°¿öµå(Keyword) |
|
ÆÄÀÏ÷ºÎ |
PDF ´Ù¿î·Îµå
|