Linux Foundation - English  Linux Foundation - Japanese  LINUXFOUNDATION.ORG | LINUX.COM | TRAINING | more Login LOGIN  Sign Up SIGN UP
Events Home

Platinum Sponsors
Hitachi LogoIBM LogoIntel LogoNEC Logo

Gold Sponsors
CE Linux Forum LogoFujitsu LogoHP LogoNovell Logo

Silver Sponsors
NTT Data LogoVA Linux Japan

Bronze Sponsors
Grape Systems LogoMiracle Linux LogoMontaVista LogoRenesas Logo

For information on sponsorship or exhibition opportunities at the Japan Linux Symposium, please contact Angela Brown at angela (at) linuxfoundation dot org.


Register for Japan Linux Symposium
Home  |   Slides  |   Sessions  |   Schedule  |   Sponsors  |   Hotel & Travel  |   FAQ

Per Backing Device Dirty Data Writeback Replaces pdflush Driven Writeback in an Attempt to Speed up this Operation

Jens Axboe, Oracle

The kernel uses pdflush threads to write out dirty data. There are, however, some problems with our current setup. pdflush has to work in non-blocking mode since it handles multiple devices, which can cause request starvation against a particular device since it cannot afford to wait for request allocation. Therehave also been reports on really fast devices out running pdflush, more than one thread would be needed to keep them running at full speed. This is not possible with the current design. We have implemented a new design for flushing dirty data, in which dirty inodes and flushing is tracked on a per-backing device basis. This reduces both locking scope and improves locality by keeping the flushing local to one (or a set) of thread(s). This solves both excessive scanning and the request starvation issue. Initial results are very enouraging, which I hope to be able to detail at the conference.