Memory Mapped Files: does it reduce Flash wear?
- From: dave@xxxxxxxxxxxxx (Dave Van den Eynde)
- Date: 4 Apr 2005 02:19:09 -0700
I'm working on a project that wants to do a large number of write
operations (journaling of application processes) on persistent memory.
The industrial environment in which our app runs demands for the shock
resistance and operating temperature range of flash-based products,
but the write-cycle wear is an issue.
In order to avoid writing to the FAT table of a CompactFlash card, I
was looking at the use of memory mapped files. My idea was to allocate
a fairly large file using CreateFileMapping and then implement a
circular buffer of some sort in our application.
The question I wanted to ask was whether a write operation to a mapped
file (i.e. writing to a memory location) invokes a write to the FAT or
the directory structure. I'm concerned about the directory entry of
the file, since time/date may be adjusted. Since no re-allocation
needs to be done the FAT should stay untouched.
I'm aware of the way "wear levelling" works on CF. I think that if a
mapped file write means a directory entry write, then the only
solution will be to write to CF sectors from our application. This is
something I would like to avoid.
Has anyone else had any similar experiences?
Thank you.
Dave
.
- Follow-Ups:
- Re: Memory Mapped Files: does it reduce Flash wear?
- From: Steve Maillet \(eMVP\)
- Re: Memory Mapped Files: does it reduce Flash wear?
- Prev by Date: Re: how to wakeup from a power save mode?
- Next by Date: IT8212 IDE Chip Problem
- Previous by thread: how to wakeup from a power save mode?
- Next by thread: Re: Memory Mapped Files: does it reduce Flash wear?
- Index(es):
Relevant Pages
|
