Abstract
While periodic checkpointing has been an important mechanism for tolerating faults in high performance computing HPC systems, it is cost-prohibitive as the HPC system approaches exascale. Applying compression techniques is one common way to mitigate such burdens by reducing the data size, but they are often found to be less effective for scientific datasets. Traditional lossless compression techniques that look for repeated patterns are ineffective for scientific data in which high-precision data is used and hence common patterns are rare to find. In this paper, we present a comparison of several lossless and lossy data compression algorithms and discuss their methodology under the exascale environment. As data volume increases, we discover an increasing trend of new domain-driven algorithms that exploit the inherent characteristics exhibited in many scientific dataset, such as relatively small changes in data values from one simulation iteration to the next or among neighboring data. In particular, significant data reduction has been observed in lossy compression. This paper also discusses how the errors introduced by lossy compressions are controlled and the tradeoffs with the compression ratio.
Highlights
The future extreme scale computing systems [13, 35] are facing several challenges in architecture, energy constraints, memory scaling, limited I/O, and scalability of software stacks
Because scientific datasets are mostly floating-point numbers, a naive use of compression algorithms can merely bring a limited improvement in terms of the amount of data reduced while bearing a high compression overhead to perform compression
This paper argues that while the traditional checkpointing continues to be a crucial mechanism to tolerate system failures in many scientific applications, it is becoming challenging in the exascale era mainly because of limited I/O scalability and associated energy cost
Summary
The future extreme scale computing systems [13, 35] are facing several challenges in architecture, energy constraints, memory scaling, limited I/O, and scalability of software stacks. Because scientific datasets are mostly floating-point numbers (in single or double precision), a naive use of compression algorithms can merely bring a limited improvement in terms of the amount of data reduced while bearing a high compression overhead to perform compression. Lossy compression on checkpointing implies several challenges for large-scale simulations, e.g., guaranteeing point-wise error bounds defined by the user, reducing a sufficiently large amount of storage space, performing compression in-situ (to reduce data movement), and taking advantages of data reduction by potentially being able to use locally-available non-volatile storage devices. Overall, applying data compression at runtime for checkpointing will become appealing to large-scale scientific simulations on future high-performance computing systems, as it can reduce storage space as well as save the energy resulted from data movement.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
