First of all, please understand my replies in the context of my original posting:
Which i quote: =============================================================== >From [EMAIL PROTECTED] Mon May 20 12:54:35 2002 Date: Sat, 18 May 2002 14:58:44 +0800 (PHT) From: Ian C. Sison <[EMAIL PROTECTED]> To: [EMAIL PROTECTED] Subject: Re: [plug] IDE vs SCSI RAID On the contrary, the main problem with IDE drives when used in server systems are - - high CPU utilization - "lesser quality components" leading to premature failure Both issues of which are addressed by SCSI-based drives. However, applying a true hardware based IDE RAID controller removes the above issues as the onboard controller relieves the CPU of handling I/O, and implementing RAID on it now truly fulfills RAID's intention of "Inexpensive" disk arrays. Those still considering SCSI based disks should evaluate their decisions again. In the long run, IDE drives configured in RAID arrays offer bigger individual disk sizes (160GB!), faster access (UDMA 133), data reliability and redundancy, as well as true cost effectiveness. =============================================================== On Sun, 19 May 2002, Rick Moen wrote: > Quoting Ian C. Sison ([EMAIL PROTECTED]): > > >>> ...faster access (UDMA 133).... > >> > >> A mirage and an irrelevancy. That's the theoretical maximum transfer > >> rate of the bus, which can never be approached even in ideal cases, > >> because data flow is limited to the physical read speed of hard drives, > > > > An issue which is just as relevant on SCSI based solutions > > No, that's _exactly_ where there's a big difference, because of > disconnected operation. When you say "UDMA 133" is "faster access", > that is completely meaningless, because only one device on the bus can > be active at a time. Demand on bus bandwidth can never be additive, > even in ideal configurations. The advantage of disconnected operation only comes in when you have more than one scsi device in the chain. IDE RAID solves this problem by having multiple IDE channels, with only one device connected to each channel. Also, By faster access i meant to compare to previous IDE technology, which did not support RAID, and was contrained to dismal data transfer rates on the wire. With UDMA66-133, the issue of data transfer no longer was in the speed of the 'wire' but in the actual data transfer from disk platters to the underlying disk electronics, which is currently the bottleneck in both scsi and ide systems, in single drive installations. > Because of SCSI disconnect, one might be potentially drawing data at the > maximum physical read speed from several drives on the chain. Thus, any > limit on bus bandwidth has some chance of being meaningful, since the > bandwidth consumption of multiple drives on the chain can be additive. Agreed. > >> I hear of some contrived test situations where it's starting > >> to be possible to saturate ATA/66, but just barely. > > > > Most Hardware IDE RAID controllers recommend one disk per ATA channel > > anyway, so it's a non-issue. > > That is _utterly_ irrelevant to my point, which is that claims that the > latest ATA extension is "faster" (because of theoretical bandwidth > maximums that can't be even remotely approached by the single active > device possible on each chain at any time) are meaningless. Thus: a > mirage and an irrelevancy. I get your point that SCSI allows you to aggregate the bandwidth consumption on a single bus. I hope you get mine when i said the latest ATA is "faster" then its previous incarnations, when the wire speed was the limiting factor. Again, I'm merely pointing out (as in my original post) that IDE RAID solutions approach the speed and reliability of SCSI based systems for a fraction of the cost. > There might be excellent reasons to deploy ATA in a given situation. > However, you happen to have cited one of the erroneous and meaningless > ones. Actually unless you intend to daisy chain 16 devices, i don't see any need to buy a more expensive scsi solution. How do you think this will stack up speed wise and cost wise? 5 SCSI disks connected to a single U140 SCSI RAID controller versus 5 IDE disks connected to a 6 channel 3ware escalade UDMA133 RAID controller ? (one disk per IDE channel) _ Philippine Linux Users Group. Web site and archives at http://plug.linux.org.ph To leave: send "unsubscribe" in the body to [EMAIL PROTECTED] To subscribe to the Linux Newbies' List: send "subscribe" in the body to [EMAIL PROTECTED]
