T.R | Title | User | Personal Name | Date | Lines |
---|
1306.1 | More info please... | NETCAD::THOMPSON | | Mon Apr 28 1997 14:49 | 28 |
|
Hi Adrian,
>>If 3 DE500 cards are plugged into the machine (eg. on the primary bus) then
>>the performance under NetWare 4.11sp2 and DE500.LAN v3.18 drops with more than
>>30 users.
How do you know the performance "drops"? Are you running benchmarks or is
it just a dropoff from a user point of view? Also, what kind of performance
do you see with less than 30 users?
Are there 10 users off each card? Are there 30 users off one card with the
other two DE500s being idle? Please give me a little more info about your
customer's setup. At some point, I would expect to see the CPU get saturated
if you have 30 clients talking to the server at once.
>>Some questions: Is this a known problem with 3 DE500s in Prioris ZX?
No.
>>Is it correct that the driver identifies the -AA and -XA?
Yes, look at the banner that appears when the driver loads.
It tells me that you have a DE500-XA loaded.
regards,
-Brenda
|
1306.2 | | NETCAD::THOMPSON | | Mon Apr 28 1997 14:52 | 9 |
|
Also, what testing have you done that indicates it's not a problem
with the HX? Did you use the same adapters, same driver, same Novell
software, same networking equipment (repeaters, hubs, whatever you
have) to prove this?
-Brenda
|
1306.3 | Further information | NNTPD::"[email protected]" | Adrian Moore | Tue Apr 29 1997 19:05 | 52 |
| Hi Brenda,
Thanks for responding.
>>>Is it correct that the driver identifies the -AA and -XA?
>Yes, look at the banner that appears when the driver loads.
>It tells me that you have a DE500-XA loaded.
This should have read "Is it correct that the driver identifies the -AA as
-XA?"
There is no part # difference, is there? To his knowledge he ordered -AA cards
but the driver reports -XA.
>Are there 10 users off each card? Are there 30 users off one card with the
>other two DE500s being idle?
It's a switched network with actually only two of the cards really being load
tested. The third card being present in the ZX seems to induce the performance
issue, whereas it doesn't on the HX.
The performance drop is reported by the client with respect to btrieve
operations and is of the order of 1:15... i.e. a similar operation on the
Prioris HX takes a minute, whereas on the Prioris ZX it takes 15 minutes.
The drives are external on the Prioris ZX, so the test with the Prioris HX
was with the same software, and the same three DE500 cards. The PCI slot
numbers were reported differently on the Prioris HX but that is all. The
HX is bridged PCI wheras the ZX is dual peer). The test took place with the
cards connected to the same network nodes... same repeaters/cabling. The
onset of the performance problem is gradual and may reflect an issue with the
ZX and bus utilisation which isn't present on the HX and isn't present if
only 2 cards are used.
I am suspecting a motherboard/compatibility issue so I will cross post in
PRIORIS. Any info you have would be welcomed, however ;] I've got the ZX
NVRAM SETTINGS.TXT file off the ECU disk for the machine but it's a bit thin
on PCI settings. This is an Intel motherboard b.t.w.
> I would expect to see the CPU get saturated
> if you have 30 clients talking to the server at once.
Maybe. Maybe the ethernet is getting saturated. The concern is that the
Prioris HX doesn't see this and that the determining factor seems to be
due to the DE500 cards. Previous cases indexed by COMET weren't adequately
documented or resolved, so I suspect this is a genuine issue.
(refs: CTI_SRC960628005967 , SRC970303005696 )
Thanks & regards
Adrian Moore
Sydney CSC
[Posted by WWW Notes gateway]
|
1306.4 | SROM info | NETCAD::THOMPSON | | Tue Apr 29 1997 20:52 | 21 |
|
Hi Adrian,
The banner for the driver is printed using the information contained
in the SROM. Thus, if the SROM says that it's a DE500-XA, that's
what gets printed in the banner. I suspect that your customer has
DE500-XAs, but would need to "dump" the SROM and visually inspect the
card to be sure.
By the way, I use other bits in the ethernet controller registers
to do the init, so I don't rely on the SROM contents to do my
initialization for the adapter. The DE500-AA and DE500-XA are
different beasts to init, so I do a chip check rather than using
the SROM info.
Right now, I can't give you any more information on your performance
problems - sounds like you're headed down the right path. Good luck!
regards,
-Brenda
|
1306.5 | Thanks heaps! | NNTPD::"[email protected]" | Adrian Moore | Wed Apr 30 1997 19:52 | 8 |
| Hi Brenda,
Thanks heaps for the braindump! ;]
Thanks & regards,
Adrian Moore
Sydney CSC
[Posted by WWW Notes gateway]
|