In the process of answering questions for others, I have tried
researching this and can't determine for sure what the answer is.
In the original presentation of Amanda
http://www.usenix.org/publications/library/proceedings/sa92/silva.pdf
(that's actually a pretty cool historical document), it clearly said
that a backup has to be completed to the holding disk before it would
then be run out to tape. It also said that if a backup was too large to
fit on the holding disk, it would be streamed to the tape sequentially,
after other backups had been completed.
Since then a lot has changed. The size of things that people are dealing
with has gotten much larger, and Amanda can now span tapes.
With that ability to span tapes, does Amanda still require a DLE to be
completed to holding disk before it will send any of it to tape? And,
does it still stream sequentially to tape if a DLE is too large for a
holding disk? And, what if you are doing server side compression? Is it
the entire DLE that gets compressed in one go? That would certainly
answer the first question.
I realize that the ultimate answer is that it is better to break up
DLE's so that you don't have any so large, thus allowing parallelism
from, say, a large SAN that in fact has multiple luns and is capable of
parallelism itself. But, knowing how Amanda is going to handle a very
large DLE does help explain options and tradeoffs.
--
---------------
Chris Hoogendyk
-
O__ ---- Systems Administrator
c/ /'_ --- Biology & Geology Departments
(*) \(*) -- 140 Morrill Science Center
~~~~~~~~~~ - University of Massachusetts, Amherst
<[email protected]>
---------------
Erdös 4
