>does the chunksize have any performance benefiot/penalty ?
No penalties I can think of at the 1 GByte level. If you went silly
and told it 100 KBytes, there is an additional 32 KByte header on each
chunk so you'd waste a tremendous amount of space, and I suppose at that
small a size there might be some speed loss as Amanda was constantly
doing the protocol and search for the next chunk.
>Or is it more a convenience (as in this case where picking the right
>chunksize will have the load indeed split over multiple disks)
The original intent was to support file systems that cannot handle files
larger than 32 KBytes (e.g. ext2 on Linux). But, as with all good ideas
:-), it turns out to have some other nice benefits, such as what you're
trying to do.
When tape chunking is done, it will also allow disk chunks that have
made it to tape to be released early (before either the dump or the tape
is completely done), which will reduce the amount of holding disk space
needed and also allow everything to flow through the holding disk, not
just the complete images that will (were estimated to) fit, which will
increase parallelism.
> Gerhard
John R. Jackson, Technical Software Specialist, [EMAIL PROTECTED]
