My newsreader is acting funny and dropping posted messages, so I
apologize if this shows up twice.
My comments on the file speed are that the times posted by other just go
to show how difficult it sometimes is to make good timing measurements.
I suspect that the wide variations being posted are in large part to
disk caching. To measure this, you should either flush the caches each
time or run them multiple times to make sure that the cache affects them
more equally.
Here is what I'd expect. The LV file I/O is a thin layer built upon the
OS file I/O. Any program using file I/O will see that smaller writes
have somewhat more overhead than a few large writes. However, at some
size, either LV or the OS will break the larger writes into smaller
ones. The file I/O functions in general will be slower to read and
write contents than making a file copy using the copy node or move node.
Sorry if I can't be more specific, but if you have a task that
seems way to slow, please send it to technical support and report a
performance problem. If we can find a better implementation, we will
try to integrate it.
Greg McKaskle
