[gclist] DSM and GC

[Patrick Panke]

---Xavier Blondel <Xavier.Blondel@inria.fr> wrote:
>
> 
> 
> >Hi everybody,
> 
> 
> Hello!
> 
> >I'm working on a project at my university, and
> >therefore have to implement a distributed garbage
> >collection algorithm that collects objects in a
> >shared memory environment.
> >
> >Ideally it should have the following prperties:
> 
> 
> As should any DGC have ;-)
> 
> [...]
> 
> >
> >I couldn't find anything appropriate online so far.
> >
> >Any tips, ideas, references ?
> >
> 
> You can have a look at Larchant:
> http://www-sor.inria.fr/projects/larchant/
> http://www-sor.inria.fr/publi/MDCSGC_ecoop98.html

Done, but seems completely unefficient.
"While theoretically complete, this strategy is
extremely inefficient. [...] We plan to implement
only the locality based heuristic..."



 
> and at PerDiS:
> http://www.perdis.esprit.ec.org/
> more precisely:
> http://www-sor.inria.fr/publi/PPCS_oopsla97.html

uses the same GC-algoritm as Larchant
 
> And if you get bored with INRIA-SOR stuff ;-), you
can still have a look at:
>
http://stork.ukc.ac.uk/computer_science/Html/Jones/gc.html
> 
> There must be something about DGC in the FAQ (even
though I did not manage
> to retrieve it...)
This was one of the first docs I read, but it does
not really explore distributed GC with replicated
Objects;
only says reference counting / listing.


Doesn't reference counting/listing involve keeping a
global count/list?
Because a reference can be destroyed on a different
computer than it was created, therefore leaving the
local counts in an inconsistent state.

Any ideas?
How do I collect replicas of the object?

Thanks,
  Patrick

_________________________________________________________
DO YOU YAHOO!?
Get your free @yahoo.com address at http://mail.yahoo.com