On 13-10-2011 1:47, Darren New wrote:
> On 10/12/2011 13:31, andrel wrote:
>> That is what 'surprises' me everytime. How can you design an OS where
>> it is
>> possible to prevent a task switch to a taskmanager. I assume it is
>> because
>> it wasn't designed but grown. Still, by now they should have solved
>> that, I
>> would assume.
>
> It's really not too hard. Modern desktop OSes suck at scheduling disk
> I/O, so if you're locked up because of disk I/O and you launch a new
> program, you're going to be locked up.

Might be, but why design it this way? Why has cntrl-alt-del not a higher 
priority than disk-IO and indeed any other program/process? (and on 
really modern machines: why not dedicate one core to the OS and the OS 
alone?)

>
> Sit down at a Linux machine with 16G of RAM. Create a directory with
> 1,000,000 one-block files, then type "sync", then type "rm -r xyz;sync"
> and watch Linux "lock up" for several minutes as anything that wants to
> touch the disk waits for the sync to finish.
>
> Same thing can happen if you have a whole ton of data paged out when you
> terminate the application, as AFAICT both Windows and Linux will happily
> page everything back in as it terminates the job.
>
>> BTW this is when reading a 1-2 million faces text file in into
>> Blender. And
>> when going into edit mode after that and...
>
> Disk I/O sucks in every modern desktop OS. Android also "locks up" on
> occastion for several seconds at a time, as some random program launches
> a check for something online when you're in the middle of typing
> something, for example.
>
> Contrast with CP/V, a mainframe OS from the late 1960s, that ran on a
> computer with *maybe* a 500KHz CPU running out of 256K of magnetic core
> that happily supported 40 or 50 users before you started to notice any
> slow-down. How? They had something like 30 or 40 different priority
> bumps depending on what you were waiting for and what woke you up.
> Waking up from waiting for free memory (i.e., someone else's page-out
> completing) vs waking up from waiting for pages to page in vs waking up
> from waiting for a directory entry to come from disk vs waking up when
> disk data has arrived vs etc all had different priority bumps, including
> on whether you're in core, out of core, finished your quantum last time,
> waited before your min-quantum expired last time, etc etc etc.

IIRC also the Amiga did not suffer from virtual lock up (nor do I 
remember it so badly from earlier version of MS-DOS).


-- 
Apparently you can afford your own dictator for less than 10 cents per 
citizen per day.

Post a reply to this message