On 25 February 2010 12:42, Ben Kloosterman <[email protected]> wrote:
>>
>  >Note that the restore from hibernation is still incredibly slow so
>  >everything that can be paged out by writing data to filesystem or
>  >evicting file cache is paged out and then the remaining used memory is
>  >compressed to write as few disk blocks as possible.
>
> Its fast enough for me for bootup and shutdown and in fact is faster than a 
> normal boot.

Still orthogonal persistence should be even faster as you load/store
only data of processes that are actually running.

Note also that the hibernation feature is as fast as it is only thanks
to paging which allows shrinking the amount of data which is going to
be saved. It does not save the whole memory and certainly does not
load the whole disk.

>  >
>  >If you wanted a system without paging it would have to read the whole
>  >disk at startup so that it never has to access it which would be
>  >useless with current disk speeds and sizes.
>
> There are plenty of systems with no paging which have no secondary storage. 
> The issue is implementing persistence with paging , though this also can be 
> done better via other methods none of which require a full read /store of 
> secondary storage.

I don't see how you implement persistence without a persistent store.
Paging is state of the art way of accessing these stores. If you have
better mechanism the I would like to know what it is.

It is certainly possible to have a system which is an isolated
appliance booted from ROM and running purely on RAM until it is
powered off and whose sore interface to the outside world are a few
buttons on its surface but a general purpose system should handle more
than that.

Thanks

Michal

------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
_______________________________________________
CapROS-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/capros-devel

Reply via email to