This message is from the T13 list server.
> Subject: RE: [t13] hmmm.. no comments?
Once upon a time I saw a fua write implementation that received the write data into
cache, flushed the whole cache, and then copied status in.
This meets "the added restriction that the data must be to the medium before the drive
can say the transaction is complete. That's it."
But then the worst case time to complete the write rises with size of cache.
Is that ok?
That implementation doesn't fit well with the sometimes popular concept of a time
limit applied per command without regard to cache history or blocks per command.
Pat LaVarre
-----Original Message-----
From: Nathan Obr [mailto:[EMAIL PROTECTED]
Sent: Thu 6/26/2003 3:44 PM
To: T13 List Server
Cc:
Subject: RE: [t13] hmmm.. no comments?
This message is from the T13 list server.
I am sending this as a test. I haven't been able to get onto the T13
List Server yet although I have been trying to reply to this thread for
a while. I apologize if I reawaken the issue. For those who were not
at the T13 meeting this week, the original issue has been resolved and
hopefully the below clarification will resolve any outstanding ones.
------------------------
[My original, long delayed email]
Everyone has made far too big of a deal out of this. First off, FUA
doesn't affect queuing or write ordering. The only difference between
FUA I/O and other types of I/O is that FUA has the added restriction
that the data must be to the medium before the drive can say the
transaction is complete. That's it. The drive may still queue the
command and handle it however it sees fit. It seems like there should
be more to it, but there isn't.
Hopefully, you can see that there is no need for concern over overlapped
LBA or affects on the queue as it doesn't affect write ordering or the
queue.
By the way, before too many people decide to jump ship for T10, you will
be interested to know that SCSI has had support for FUA for quite some
time (See SBC). Many hard drive manufacturers shouldn't have a problem
implementing this as much of it can be borrowed from their existing SCSI
lines.
Nathan
