Re: [Gems-users] Having trouble getting LogTM running

[Dan Gibson <degibson@xxxxxxxx>]

First of all, I cannot comment on the naked scripts since I haven't used 
them myself. I generated many of our internal checkpoints, the 
old-fashioned manual way.
That said, I advise you not to run two instances at the same time. The 
limiting factor in the OS install is memory, and unless you have a large 
main memory on your host, you'll probably slow things down considerably 
if you run two instances.
When I actually installed the OS, I installed on a one-processor machine 
with a small memory, then booted the disk images on different machine 
configurations to make 2, 4, 8-processor (etc) checkpoints. As I recall, 
the OS install (both disks, Solaris Express) took around 8 hours or so 
on an Opteron 248 w/ 16 GB of memory, so it could take considerably 
longer than that to do so on a 32-processor machine. Just booting a 
32-processor machine (note, after OS install) can take up to four hours 
according to my notes.
Its fairly common to reduce the target's CPU frequency (or at least what 
Simics pretends it is) to relatively speed up I/O operations. You can do 
so by changing TARGET_CPU_FREQ_MHZ int system.conf to, say, 10-20. This 
won't affect the statistics reported by Ruby, aside from changing 
relative frequency of interrupts and I/O events.
Lastly, I haven't installed Solaris 10 before, but the install process 
for Solaris Express is supposedly similar and the only progress bars I 
recall come from the individual package installs, after machine 
identification. Judging by your output (in the previous message) the 
installer hasn't yet reached machine ID.
Regards,
Dan

Jim Leek wrote:
> On Thu, Feb 28, 2008 at 8:41 AM, Dan Gibson <degibson@xxxxxxxx 
> <mailto:degibson@xxxxxxxx>> wrote:
>     Installation of the OS can take a LONG LONG time to run. Having 32
>     processors and 64 GB of memory will only slow things down. If you're
>     concerned about excess runtime, I'd recommend trying to make a 2-8
>     processor checkpoint with 4 GB of RAM (I believe the LogTM papers used
>     machines with 4 GB, though an author might correct me on that).
>
>
> 64GB of RAM did seem excessive, but I didn't find how much RAM was 
> used in the papers I read.  I guess I'll try another install  as you 
> suggest.  I have to processors on this machine, so I don't think it 
> should be a problem to run another run of the script on a smaller target.
> How long is a LONG time?  How many days are we talking about here?  I 
> would have expected, at the very least, some sort of progress bar from 
> Solaris.  I haven't seen any new output in 13 hours, but I've also 
> never installed Solaris before, so maybe that's normal?
> Jim
> ------------------------------------------------------------------------
>
> _______________________________________________
> Gems-users mailing list
> Gems-users@xxxxxxxxxxx
> https://lists.cs.wisc.edu/mailman/listinfo/gems-users
> Use Google to search the GEMS Users mailing list by adding "site:https://lists.cs.wisc.edu/archive/gems-users/"; to your search.
>
>   
--
http://www.cs.wisc.edu/~gibson [esc]:wq!