Article: AMD's Mobile Strategy
By: Ricardo B (ricardo.b.delete@this.xxxxx.xxxx<), December 22, 2011 8:36 am
Room: Moderated Discussions
anon (anon@anon.com) on 12/21/11 wrote:
---------------------------
>Oh, OK. What kind of data integrity operations is it doing that require a significant
>amount of bandwidth in random writes?
I don't know how Windows and NTFS behave but there may be some interactions at hand that cause problems.
A good example is Linux's ext3.
Under the default setting (ordered data mode), before commiting a segment of journal (every 5 seconds), ext3 commits all the related data.
This is very feature to have for applications that don't sync.
But it also lead to sync being a very expensive operation on ext3, not only from a thread/application POV but even from system POV.
---------------------------
>Oh, OK. What kind of data integrity operations is it doing that require a significant
>amount of bandwidth in random writes?
I don't know how Windows and NTFS behave but there may be some interactions at hand that cause problems.
A good example is Linux's ext3.
Under the default setting (ordered data mode), before commiting a segment of journal (every 5 seconds), ext3 commits all the related data.
This is very feature to have for applications that don't sync.
But it also lead to sync being a very expensive operation on ext3, not only from a thread/application POV but even from system POV.
