[gclist] Finalizers & Reference counting.

Charles Fiterman cef@geodesic.com
Tue, 27 Aug 2002 05:14:46 -0500


At 06:33 PM 8/26/2002 -0700, you wrote:
> > -----Original Message-----
> > From: Charles Fiterman [mailto:cef@geodesic.com]
> > As systems expand the time taken by collection becomes a
> > fatal problem. One
> > company had contracts that said 90% of transactions must be done in 1
> > second and the rest in 10 seconds. When the system expanded
> > past 1 GB they
> > started to fail due to garbage collection time. Hacking got
> > them past the
> > problem but at some point such guarantees must fail. At some point
> > collection times must exceed any given limit.
>You are assuming that memory usage increases without bounds, but processor 
>speed, and the number of processors, remain fixed.  In my experience 
>that's not the general trend.  In the early 80s the general rule of thumb 
>was to configure a machine with about as much memory as it could touch in 
>a second.  My current home PC has less memory than that.  Multiprocessors 
>are more and more common, and GC scales well.

Actually I was assuming the number of processors increased with memory but 
that they couldn't fully divide collection. That is each processor would do 
part of collection but they would have to cross talk and the cross talk 
would increase without bounds.