Thanks Franky. In terms of performance, I don't think AFD or ASMLib has any
advantage over not using them. ASMLib offers management convenience (or
INconvenience to some people). AFD offers write protection from non-Oracle
programs. Our shop has stopped using ASMLib for some years. Now it's time for
us to formally install 12c, our team is debating whether to embrace its
successor AFD. We don't worry about non-Oracle writes since it's extremely
unlikely for us. We kind of worry about the kernel module. Now Red Hat kernel
update includes ASMLib (I think), so it probably will include AFD as well. If
not, we would have to manually compile the AFD from source every time the
server team plans a kernel update.
I'd love to know what the trend is. I wish there was a poll on whether
ASMLib/AFD is or will be used.
Another reason I worry about ASMLib/AFD is bugs. Using a technology without an
additional layer is my preference, unless of course that layer offers benefit
we do need but can't easily implement.
Yong
----- Original message ----
I never really tested the performance, but it looks nice to use. We have a--
small 3 node cluster running with AFD, so I still don't know what to expect
for large environments, since our core here runs on Exadata.
I think there is plenty of reasons to use it.
When the filter is enabled you can't mess up the disks either trying a
simple echo or dd command, even as root.
Franky Weber Faust
