Re: AW: Reconstruct aggregates on diskpools

Steffan, 

I don't really know if moth-eaten aggregates have a performance penalty;
maybe someone from development can provide an answer.

My gut feeling is that there in NOT any penalty.  My guess is that each 
aggregate
can be accessed directly for each file that is still valid.  I don't think 
there is
any difference in the resote of one file from an agg. made today and from
one made a year ago.  Again, that is just a gut feeling, I haven't tried to 
measure
anything.

-- 
--------------------------
--------------------------
Bill Colwell
Bill Colwell
C. S. Draper Lab
Cambridge, Ma.
bcolwell AT draper DOT com
--------------------------
 
 
In <96A7493F33B7D4119EA80001FAD4B0AF01364474@SZENT008>, on 11/13/01 
   at 12:31 PM, Stefan Holzwarth <stefan.holzwarth AT ZENTRALE.ADAC DOT DE> 
said:

>Hi Bill, thanks for your input. 

>What do you think about aggregates on diskpools over a long time? 
>Is there much performance penalty due to inefficent aggregates with many
>gaps inside?

>Regards, Stefan Holzwarth


>> -----Ursprüngliche Nachricht-----
>> Von: Bill Colwell [mailto:bcolwell AT DRAPER DOT COM]
>> Gesendet: Dienstag, 13. November 2001 15:26
>> An: ADSM-L AT VM.MARIST DOT EDU
>> Betreff: Re: Reconstruct aggregates on diskpools
>> 
>> 
>> Steffan,
>> 
>> Reconstruction is only done when the input and output pools are both
>> sequential.  If you have 4.2 you can migrate the diskpool 
>> down to 0 to a tapepool,
>> then do a move data with 'reconstruct=yes' to another tape in 
>> the pool,
>> then do move data back to the diskpool.  If you aren't 
>> running 4.2 this is much harder
>> since only reclaim will do reconstruction but the tape must 
>> be full to be a
>> reclaim candidate.
>> 
>> 
>> --
>> --------------------------
>> Bill Colwell
>> C. S. Draper Lab
>> Cambridge, Ma.
>> bcolwell AT draper DOT com
>> --------------------------
>> 
>> 
>> In <96A7493F33B7D4119EA80001FAD4B0AF01364468@SZENT008>, on 11/13/01
>>    at 09:25 AM, Stefan Holzwarth 
>> <stefan.holzwarth AT ZENTRALE.ADAC DOT DE> said:
>> 
>> >Does anyone know, whether there is some kind of 
>> reconstruction of aggregates
>> >within a diskpool? Since many of us use a diskpool for 
>> directories with no
>> >tape as migration target,
>> >how "good" are the aggregates after a longer period?
>> >I'm no satisfied with the performance of the copy 
>> dirbackuppool to tape = ~
>> >1 hour
>> >Maybe thats a reason..
>> 
>> >Regards, Stefan Holzwarth
>>