There used to be an official Posting of the SAH Administration several month ago :

Ideal Cache Size : 24-48hrs max.
(Results get fully accounted into Database, since most Results return within that
Timeframe)

Max. desireable Cache : ~7 days
(some of the Results are obsolete already, but since some never return at all, still
limited useful for science. Most of those old Results also serve "Database Integrity"
checks)

Greater than 1 week (e.g. upto 1 month) :
Almost fully redundant.
No scientific use, almost all Results have the 'verified' Tag already in the Database. Too
old even for Integrity checks, and increasing the stats counter only ...

This is what I recall, and the average Return rate will sure have slightly dropped again
within the last months when the statement was made.

The idea of big, long running local Caches was fully supported a long time ago, when
average Return rates were alot lower (and the gross computing power to SAH much smaller as
well); and the Server often was overloaded, dropping lots of connections on a
regular/normal basis.

But these times are definitely over.

I therefor limited my Caches to max. ~24hrs long ago, since one has to add the actual
computing time, and the delay to the next Cache Flush onto it as well (e.g. a machine
working 8hrs will return a Result it after a total of 32hrs minimum, no delays included)
With the Congent Uplink mostly working flawless, I think that's a good technique.

So IMHO, 'getting them back home' should happen ideally a rough 36hrs after reception.

Greetings
FalconFly

Webmaster
http://www.falconfly.de
3dfx Archive
====================
Forum : http://www.falconfly-central.de