Re: [Pan-users] Disk dowload slowdown

[David Kelly]

On Oct 12, 2006, at 7:49 PM, fred wrote:

> David Kelly wrote:
> > Doesn't matter if par2 is cranking its heart out at 25 MB/sec at the
> > same time as pan is downloading. Pan actually slows par2 which will
> > usually run at disk speed if it has the drive to itself.
>
> I neglected to mention that this is when it is occurring for me as  
> well.

Trust my trimming this reply has correctly connected Fred/Lazlow's  
reply to my prior statement for which he intended.

As for slowdown believe my statement was in opposition to Fred/ 
Lazlow's original. That in my case pan does not slow during download  
due to disk activity. But if pan is writing 500 kbytes/sec to the  
drive at the same time par2 is crunching then par2 (not pan) runs  
slower than it would if pan was not also writing to the same  
mechanism. Par2 runs at 20 to 25 MB/sec when it would be over 40 by  
itself.

By itself par2 will consume 100% of the drive bandwidth.

Pan is capable of consuming 100% of the drive's bandwidth when  
articles are moved out of cache into the final destination. If pan  
and par2 get throttled to about half of what is available when they  
contend for 100% at the same time. As one might expect the total  
drive bandwidth is less when two tasks are greedy than if one greedy  
task has the drive all to itself.

Two tasks contending for all the drive bandwidth they can get drop  
the total throughput to 10 to 20 MB/sec when a single task could have  
sustained 40 to 50. A large file copy usually runs just over 20 MB/ 
sec, 10 reading, 10 writing. Matters a lot how far apart the two sets  
of data are on the disk as to how far the heads have to move between  
tasks. When copying the fs is free to choose empty blocks close by.

--
David Kelly N4HHE, address@hidden
========================================================================
Whom computers would destroy, they must first drive mad.