[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Torsion-dev] Storage -- segments & pages
From: |
Davison Avery |
Subject: |
[Torsion-dev] Storage -- segments & pages |
Date: |
Sun, 5 Feb 2006 02:11:02 -0500 |
User-agent: |
Mutt/1.5.11 |
Dan,
Great work on TorsionOS. I was reading your wiki page about the
page-based storage system.
I have a few preliminary questions, without having yet gone through your
source (which I plan to do):
1. When you are checkpointing, I want to confirm that you only commit
the dirty pages that correspond to a specific point in time, since a
checkpoint (if I am not mistaken) should correspond to the state of
virtual memory at a specific instant in time.
2. With regards to your disk layout, why did you choose to use segments
and pages? Was it because pages seem to be more low-level and are more
dependent on block size, while segments were meant to be more abstract?
I don't see the reason why the disk layout couldn't have stayed at the
page level.
EDIT: After reading your paragraph on Checkpointing, it seems to me that
you are treating segments as a way to track disk utilization only; while
pages are specifically virtual address space related only. Based on this
understanding, would it be correct to indicate that segments would (for
a 32bit system) be 4GB in size (not including supeblock sizes), and that
at minimum the whole system would require two segments -- one for
checkpointing all dirty pages, and another segment to hold the last
successful checkpointed image?
Thanks!
[Prev in Thread] |
Current Thread |
[Next in Thread] |
- [Torsion-dev] Storage -- segments & pages,
Davison Avery <=