The schedule is still fluid at the moment as the final approved sessions are added. We appreciate your patience.
Back To Schedule
Tuesday, May 20 • 3:00pm - 3:50pm
Improvements on Kdump Scalability Issues for Terabyte-Scale Memory System - Hatayama Daisuke, Fujitsu Limited

Sign up or log in to save this to your schedule, view media, leave feedback and see who's attending!

There had been several scalability issues on kdump framework with
recently emerging terabyte-scale memory systems. Due to the issues,
even partial crash dump required over 1 hour. Recently, there are
several improvements for the issues: we implemented zero copy with
mmap() and optimized makedumpfile to address bottlekneck of reading
crash dump from memory, and we now support parallel compression by
supporting fast compression format and by removing 1-CPU limitation on
kdump 2nd kernel. Of course, the improvements have also good effect on
systems with typical amount of memory.

This talk explains these improvements for the kdump scalability
issues, then explains also how to use them on kexec-tools on recent
fedora and RHEL7 systems and finally shows benchmark on them.



Software Engineer, Fujitsu Limited
HATAYAMA Daisuke has been working on support job for mission critical fields like financial service for 5 years, specifically on crash dump system and debugging tools. He has also committed several open source communities relevant to kdump framework.

Tuesday May 20, 2014 3:00pm - 3:50pm

Attendees (0)