GF> On 26 Dec 97 07:56am, Chris Holten wrote to Wes Newell:
 GF> You should have seen those older clients REALLY scream when
 GF> they insisted on adding a SECOND IDE drive to the server!
 GF> You KNOW what kind of a performance hit they took then.
 MB> Yes, that's a clear mistake to do.
 WN> OK, I'll bite. I added a second IDE drive (WD 1.6gig)
 WN> to our 3.12 server and didn't notice any slowdown. I
 WN> did add it to the second IDE channel though, so there's
 WN> only one drive per channel. Perhaps this is why there
 WN> wasn't a decrease in performance?
 CH> I wouldn't worry at all about using EIDE on a novell server. In
 CH> most typical Novell file/print server 10bt ethernet setups
 CH> using 1997 EIDE hardware technology, it will give you just as
 CH> good a performance as any SCSI setup will as with any 100mhz or
 CH> better pentium and the amount of RAM Novell recommends, will
 CH> -easly- saturate a 10bt bandwith.  If you were to purchase a
 CH> good busmastering SCSI setup with a decent 6.4 gig SCSI drive,
 CH> you would spend at least $400-$600 more than you will on an IDE
 CH> setup. SCSI is nice, and if you ever go to 100 base T, you
 CH> might find a busmastering SCSI setup faster. It's up to you to
 CH> evaluate. Even with 100 base T, if your network is a 25 user or
 CH> less, you are probably wasting $400 on a SCSI Novell server. If
 CH> you want to increase.
 GF> And here is where you are totally wrong.  I suggest you upgrade your
 GF> benchmarking software and try it again.
 GF> I will happily stand by my claims and enjoy the fact that my clients are
 GF> getting the performance and reliability they paid for.  If you like
 GF> going back and explaining to your clients why they have problems (both
 GF> efficiency-wise and hardware-wise) periodically and quite a bit more
 GF> often in your "bare minimum" requirements system, that's your
George, you and Mike Bilow can't even make your theories about SCSI vs modern 
Busmastering IDE setups work on paper let alone with any benchmarks on a 10bt 
network using modern EIDE hardware and Pentium class computer chipsets. No 
doubt SCSI does make the Vendor a bit more profit and there is some kind of 
"down the road" justification for the client, but that wasn't the point of 
the original message. They guy already had a 6.4 gig WD Eide drive. You and 
mike tried to make him think that he couldn't/shouldn't use that, instead he 
oughta to go get a SCSI setup to gain 0 (Zero) increase in performance on his 
novell server, cause according to you two, you just always have to have SCSI 
on a novell server no matter what the server is being used for, how many 
users it has and whether it is 10bt or 100bt is irrelevant. (No body ever got 
fired by buying IBM either). Your comments, experieces on the matter were 
*obviously* not based on modern 1997 EIDE setups with moder pentium class 
chipset/bios or benchmarks on a 10bt ethernet. The type of performance 
example cited seems to me to be based more on 386 and early 486 
IDE/Bios/Chipset  technology. No doubt 3-10 years ago, SCSI could be 
significantly faster in a dedicated network file/print server environment and 
if that was the kind if equipment being used, you are spot on, especailly 
considering that back then memory was much more expensive and dedicated 
novell servers usually were not over endowed with RAM memory. That's all 
changed.
Modern EIDE setup will indeed much much more than saturate the bandwidth of a 
10BT ethernet cable. (10-18,000kbs for EIDE vs at the very most 550kbs for 
10bt ethernet). 
In view of those througputs, if you can only realistically poke at most 
550kps through a 10bt ethernet cable, how can you possibly have the eide use 
anywhere near 30-50% of the CPU just to retreive data off the hard drive. 
What is that other 50% CPU capacity going to do?
I think what you and mike are trying to relate to is what you have read about 
 IDE vs SCSI performance on a local desktop system an then try to apply that 
to a dedicated Novell file/print server (the case of the original question in 
this thread). Since the particular dedicated Novel File/Print servers 
communication to the outside world is through was through 10bt ethernet cable 
and not from a local keyboard, video, or a network multiuser application as 
something like Unix is typically used for, the whole crux of a novell 
file/print servers performance is keeping the data at the maximum rate that 
can be transmitted through a 10bt cable. It doesn't matter if the CPU is at 
80% utilization to do that or at 10% utilization. All that matters is that to 
keep the bandwidth saturated, that the demand on the CPU not be over 100%. 
Doesn't matter if you have 1 user or 10000 users. If the bandwidth is 
saturated, that's the best you are going to do. Period. Anymore CPU/Drive 
throughput capacity than that needed to reach much more than 100% bandwidth 
is unecessary. What you are and Mike are saying is that to have 2-5 times or 
-more- the ability to saturate 10bt bandwidth is not good enough if it is a 
busmastering IDE drive, cause everyone knows that busmastering EIDE drives, 
although they are as fast in througput are not -quite- as good at 
busmastering as a top of the line PCI SCSI controller. From your rules of 
thumb and postulation about SCSI, You need to have 5-15 times that ability. 
That's non-sensical and a theory that won't even work on paper, let alone the 
real world (And it doesn't...not even close).
--- Maximus/NT 3.01b1
---------------