OK
Now I have one serious question for somebody out there.
I have read review after review on IDE and SATA with RAID.
But one real element is always missing.
SCSI!!!
Now there have been a few comparisons of SCSI but they were really RAID 0 and 1.
Granted when doing apples to apples this is cool.
But what do we truly run with RAID systems??
Here at work I run all because I am a senior QA in my company. We writre backup software so I use RAID controllers all of the time. Mostly RAID 0 and 1. And this is fine because of the test regimen I am running. And all or most RAID 0 and 1 are not all that fast.
But RAID 5 and 3 although 5 is preferred, (something about parity seems important. I have seen and had tremendous throughput in RAID 5.
For example when at IBM I had a server (dual P3 Xeon 667) with a RAID 5 of 10 18 GB Seagate Barracudas (7200) rpm drives on a IBM ServeRAID 3L. This was my image server. When re-imaging my test rack of 20 plus systems at a time (each image was anywhere from 500 MB to 1.5 GB). I had a sustained throughput at the client of 280 - 340 MB per minute. I could have never sustained such a rate on any IDE subsystem. So do the math 20 x 310 MB / min??? That is some real disk IO going on.
In my testing here at Xpoint I have test my own SCSI 160 system non RAID. Just imaging to same drive different partition. And imaging to seperate hard drive.
Going from HD0 to HD1 backup MB/Min 458
Going from HD1 to HD0 restore MB/Min 956
Going from HD0 to HD0 backup MB/Min 335
Going from HD0 to HD0 restore MB/Min 714
IDE performance was about half of this. I have had up to 1.0 GB / Min performance with just SCSI 160. And yet I have read review after review that SCSI is dead.
My, my, my, if it were not for the outright cost of SCSI I am sure we would all have it. Now this was just a smal example of what I have seen between SCSI and IDE. There truly is no comparison of SCSI and IDE the performance on IDE will never get there. Too much overhead with the CPU. Only 2 devices per channel, etc...
Now a true SCSI 320 on a controller with 256 MB cache, running 10 - 14 15,000 RPM drives at RAID 3 and RAID 5. We would be talking a real comparison. Because as we know SCSI RAID is all hardware, and it's processor does all of the work so the CPU does not have too.
Then let us see who is king of the HDD and Controller world. IDE?, SATA? or SCSI?
My bet is SCSI, I have run SCSI systems for years now and hands down they are untouchable.
Pound for Pound, dollar for dollar, I will agre you can't beat IDE and SATA.
What is the price for true real performance?
Look at the server world and big workstations, what are they running? SCSI 320 and most all of them offer RAID 5 etc...
So please will someone explain to me the dilehma.
thanks
Duane
Now I have one serious question for somebody out there.
I have read review after review on IDE and SATA with RAID.
But one real element is always missing.
SCSI!!!
Now there have been a few comparisons of SCSI but they were really RAID 0 and 1.
Granted when doing apples to apples this is cool.
But what do we truly run with RAID systems??
Here at work I run all because I am a senior QA in my company. We writre backup software so I use RAID controllers all of the time. Mostly RAID 0 and 1. And this is fine because of the test regimen I am running. And all or most RAID 0 and 1 are not all that fast.
But RAID 5 and 3 although 5 is preferred, (something about parity seems important. I have seen and had tremendous throughput in RAID 5.
For example when at IBM I had a server (dual P3 Xeon 667) with a RAID 5 of 10 18 GB Seagate Barracudas (7200) rpm drives on a IBM ServeRAID 3L. This was my image server. When re-imaging my test rack of 20 plus systems at a time (each image was anywhere from 500 MB to 1.5 GB). I had a sustained throughput at the client of 280 - 340 MB per minute. I could have never sustained such a rate on any IDE subsystem. So do the math 20 x 310 MB / min??? That is some real disk IO going on.
In my testing here at Xpoint I have test my own SCSI 160 system non RAID. Just imaging to same drive different partition. And imaging to seperate hard drive.
Going from HD0 to HD1 backup MB/Min 458
Going from HD1 to HD0 restore MB/Min 956
Going from HD0 to HD0 backup MB/Min 335
Going from HD0 to HD0 restore MB/Min 714
IDE performance was about half of this. I have had up to 1.0 GB / Min performance with just SCSI 160. And yet I have read review after review that SCSI is dead.
My, my, my, if it were not for the outright cost of SCSI I am sure we would all have it. Now this was just a smal example of what I have seen between SCSI and IDE. There truly is no comparison of SCSI and IDE the performance on IDE will never get there. Too much overhead with the CPU. Only 2 devices per channel, etc...
Now a true SCSI 320 on a controller with 256 MB cache, running 10 - 14 15,000 RPM drives at RAID 3 and RAID 5. We would be talking a real comparison. Because as we know SCSI RAID is all hardware, and it's processor does all of the work so the CPU does not have too.
Then let us see who is king of the HDD and Controller world. IDE?, SATA? or SCSI?
My bet is SCSI, I have run SCSI systems for years now and hands down they are untouchable.
Pound for Pound, dollar for dollar, I will agre you can't beat IDE and SATA.
What is the price for true real performance?
Look at the server world and big workstations, what are they running? SCSI 320 and most all of them offer RAID 5 etc...
So please will someone explain to me the dilehma.
thanks
Duane