On Sat, Jan 27, 2001 at 05:40:38PM +1100, Glenn McGrath wrote: > Nathan E Norman wrote: > > > > Yeah right, I'm going to build a high performance server on IDE > > <snicker> > > > It is an undisputable fact that IDE gives better performance per $ than > SCSI. Per drive, yes. Per channel, I'm not sure I agree. I can only put two drives on an ide channel, and these drives place a burden on my cpu. For my personal desktop, that's ok. I don't accept this for a server ... Your statement still doesn't mean that ide drives can perform as well as scsi drives ;) > You realise the speed of a drive is primarily dependent on its RPM's not > its interface (or its pricetag). You don't think quality plays a factor? Can you hot swap ide drives? How many 10000 rpm ide drives are out there? I will grant you that UDMA has closed the performance gap. Hopefully this trend will continue. > > > If you have the cash to keep spare HW RAID controllers onsite, then you've > > > probably got the cash to setup duplicate servers. If you have duplicate > > > servers, you don't need RAID because you already have overall system layer > > > redundancy, so you don't need RAID. A solution in search of a problem. > > > > This does not follow. If I've got the money to keep a $1000 raid card > > as a spare I've got the money to keep a $5000 server on the network > > (which I may or may not own; more expenses), and I've got to keep the > > data synced? Yi. > > > Distributed computing will always be better than trying to cram all the > power into one machine. Agreed. However, businesses have budgets. It's not realistic to expect that every server will consist of at least two machines shring the load (if that's how it works where you are, are they hiring?) > Throwing money at a problem to avoid a technical solution demonstrates > poor managment and engineering skills. Thank you. I am of the opinion that dong everything on the cheap demonstrates a lack of management skills (though perhaps not technical skills :) (and yes I've cobbled together plenty of machines for production environments against my better judgement). Buying more expensive components does not mean that money is being "thrown" at the problem (though the it world is bad about this, I agree). > > > The only attempts to explain why HW RAID is better revolve around nonsense > > > like "its not important unless you spend extra money" or something. > > > > Uh huh ... you run a lot of raid 5? > > > Software raid is faster than hardware raid, software raid offloads the > proccessing to the CPU, whereas hardware raid does it onboard which is > why hardware raid is slower and more expensive. I agree it's more expensive. That doesn't make it "bad". There's a difference between "throwing money at a problem" and spending a little more so you can sleep at night and spend time with your kids. YMMV. Where are all the benchmarks which say one is faster than the other? I'm not comfortable with such blanket statements. > Ever seen a hardware raid card with as much power as a cpu ? Yeah, so? Perhaps it's more cost effective for that cpu to handle i/o tasks so the cpu(s) can do the other things i bought them for. Often it's some sort of RISC processor. > If you worried about overloading you system get a MP mobo. Got one, thanks :) > Most of these opinions i gathered when i was on the linux-raid mailing > list, you might want to checkout the mail archive. Ah, those are opinions :) Look, if you like software raid, more power to you. I've used it, it's cool, it has its uses. Hardware raid also works, is cool, and has its uses. I've solved problems with hardware raid that at the time could not be solved using software raid (like raid 0 on a bunch of big fast disks for a news server). I will admit that I'm probably thinking of higher end, mission critical stuff. One size doesn't fit all, which I thought was Adam's original point as well. Cheers, -- Nathan Norman - Staff Engineer | A good plan today is better Micromuse Inc. | than a perfect plan tomorrow. mailto:nnorman@micromuse.com | -- Patton
Attachment:
pgpE7RSPX7UUF.pgp
Description: PGP signature