Date: Thu, 28 Sep 2000 23:09:45 -0500 From: David Kelly <dkelly@hiwaay.net> To: chat@FreeBSD.ORG Subject: Re: CD writers - recommendations Message-ID: <200009290409.e8T49ja38223@nospam.hiwaay.net> In-Reply-To: Message from John Galt <galt@inconnu.isu.edu> of "Thu, 28 Sep 2000 21:29:16 MDT." <Pine.LNX.4.21.0009282041540.6683-100000@inconnu.isu.edu>
next in thread | previous in thread | raw e-mail | index | archive | help
John Galt writes: > On 28 Sep 2000, Dag-Erling Smorgrav wrote: > > > The reason why ATAPI CD-ROM burners are so sensitive to CPU load is > > that they don't support disconnection or tagged queueing, so commands > > must be sent sequentially and the CPU must wait for them to complete. > > So IOW IDE devices throw an interrupt whenever they're operating, SCSI > devices throw an interrupt only whenever the host adapter is starving for > commands. Sounds exactly like what I wsas saying in the first place, just > using proper terminology... I just now kicked off bonnie on my Symbios 875 UWSCSI card and IBM fast-wide SCSI drive. Initially saw 550 IRQ/sec on sym0 for 160 tps in systat -v, since then there has been at least 50% more IRQ's than t's. On another machine with UDMA100 drive on UDMA33 interface I see exactly 1:1 IRQ to transfer ratio. Yet that really has nothing to do with it at all. Only pointing out John's statements are made without corraborating evidence. The whole point of tagged queueing is the device can respond to requests in the easiest order, not the requested order? Isn't that the whole point of queuing 128 pending commands to the drive, that it can optimize its seeks? So what does that have to do with a CD-R? Since when is there anything out of order that it should be allowed to do while cutting a disc? Agreed that you don't want the source to run off doing something else for so long that the CD-R starves. But that is a real time issue. Actually its an issue that IDE drives serve better than SCSI as in a real time system "fast" is not as important as "predictable." I said it before, will say it again. IDE devices were designed to be cheap, and a good number of IDE devices live down to that expectation. But they don't *have* to live down to it. From what I've seen with my 15G IBM UDMA100 drives I probably won't be buying SCSI HD's for a while. I'm not building high volume internet servers and don't know if the thruput I've observed scales. As for CD-R and tape drives, I still believe the best devices have SCSI interfaces but suspect they will soon migrate to Firewire. -- David Kelly N4HHE, dkelly@hiwaay.net ===================================================================== The human mind ordinarily operates at only ten percent of its capacity -- the rest is overhead for the operating system. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-chat" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200009290409.e8T49ja38223>