Re: Xeon X3450 4 core speed step compatible with SCO 5.0.7

Bela Lubkin wrote:
Steve M. Fabac, Jr. wrote:

I'm loading SCO 5.0.7 on a Dell T310 with X3450 Xeon cpu.

This cpu has Intel Speed step technology that "with a supported
OS" will dynamically adjust the CPU speed as needed under
load.

I've a gut feeling that "supported OS" does not include SCO
5.0.7.

Probably right. Turn off speedstep and any other such things, give OSR5
a CPU that runs at a steady rate no matter what.

What I am seeing with hw is that the CPU are reported as:

There are 4 CPUs on this system.
Maximum: 10
Configured: 4
Active: 4

CPU 1 performs like a 1478Mhz Intel Pentium greater than III

Pay no attention to these numbers. It uses a single very crude
benchmark which is some sort of small tight CPU loop. It's extremely
dependent on CPU design details. If `hw` knows what kind of CPU you
have, it has a table of constant factors telling it how that kind of CPU
performs, and it'll probably come out with a number somewhat similar to
your clock speed. In this case it *doesn't* know the CPU type (and why
/ how would OSR507, released in 2003, know about a Nehalem Xeon released
in 2009?). It uses some goofy constant factor and gets whacky results.

The above is with two of the four cores of the X3450 enabled in
BIOS.

test_speed executes find / -depth -print | cpio -oca | cpio -icvmtn > /tmp/BACKLIST

# time /u/migrate/test_speed

real 4m19.71s
user 1m12.72s
sys 1m39.55s

This test is completely dominated by disk speed. It's useless for
benchmarking CPUs.

But not completely. The user/sys numbers are vaguely interesting, as we
see below.

I tried disabling all cores and "virtual processor" so that only
one CPU shows up and still the speed is listed as 1478 MHz (1.47Ghz)

I tried enabling all cores and disabling "virtual processor"
(no hyper thereding) and the hw then reports:

# hw -v | grep performs
CPU 1 performs like a 1276Mhz Intel Pentium greater than III
CPU 2 performs like a 1276Mhz Intel Pentium greater than III
CPU 3 performs like a 1276Mhz Intel Pentium greater than III
CPU 4 performs like a 1276Mhz Intel Pentium greater than III

X3450 probably supports the new "Turbo" mode where if less cores are
enabled, the remaining cores run at a higher clock rate. Let's see:
1478 / 1276 * 2.66GHz = 3.08GHz, is that the top "Turbo" speed of an
X3450?

CPU 1 performs like a 1478Mhz Intel Pentium greater than III

# time /u/migrate/test_speed

real 4m19.83s
user 1m11.07s
sys 1m39.04s

User/sys are the same as before. That implies that when actually doing
work, a single core in 4-core mode runs as fast as a single core in
2-core mode. "Turbo" both speeds up the remaining cores when some are
disabled; and speeds up cores when they're busy. In some models the top
speeds are the same for both even if the not-busy speeds are different;
this may be one.

On my office server with 5.0.7 and two single core Xeon CPUs
with Hyper threading I see:

There are 4 CPUs on this system.
Maximum: 10
Configured: 4
Active: 4

CPU 1 performs like a 2656Mhz Intel Xeon

"Xeon" -- that means `hw` thinks it knows this CPU type and therefore
has a sensible divisor constant for it.

...
Next, I try two cores with HT and two console screens running
/usr/games/worms -R Worms really ties up CPU.

# time /u/migrate/test_speed
real 4m25.95s
user 1m2.57s
sys 1m30.08s

Now see, this is interesting. This is the same as the first config,
except running `worms`. Notice that "user" and "sys" time are *lower*
by about 8.5 and 9s. What's that about?

Well, OSR5 time accounting is pretty sloppy. It's done in terms of
1/100 second chunks by statistical sampling. It's mostly right if just
one process is running. With more, it depends a lot on the fine scale
behavior of each process. `worms` runs flat out continuously, as much
CPU as the kernel will give it. Your `find | cpio | cpio` pipeline is
going to be very stuttery since it can only do work when the disk has
fed the input end of the pipeline. Also, the work it does when a disk
block comes in is probably far shorter than 10ms (1 100Hz time span).
So its "cross-section" for being noticed by the statistical sampler is
small, it's likely to be missed a lot. `worms` is probably being
charged extra; fortunately nobody is paying $$/CPU second for any of
this...

mpstat with only two console screens running worms -R

| |
| system | user status sys cs int tr |
| ------------- ------ --- -- --- -- |
| +--------------------+ |
| CPU1 |||||||||||||||||||| | ACTIVE 597026253 117 1 |
| +--------------------+ |
| +--------------------+ |
| CPU2 | | ACTIVE 97 16 109 0 |
| +--------------------+ |
| +--------------------+ |
| CPU3 | | ACTIVE 54 2 102 0 |
| +--------------------+ |
| +--------------------+ |
| CPU4 | | ACTIVE85723852738 26470 1 |
| +--------------------+ |

Looks like only 1 CPU lit up, I'd expect at least 2. Also the
sys/cs/int stats for CPU4 look crazy.

Change BIOS and switch from MAXIMUM PERFORMANCE to APC
...
Odd: hw hangs up under the above test conditions.

Run `hw -?` to see all the report types it supports. When it lists
something like "ram, mem, memory" it's offering you 3 aliases for the
same report. The OSR506 version I'm looking at has 11 reports. Run
each individually (`hw -r mem`, etc.), you should be able to pick out
which one is hanging.

Ok, now that you have determined that it's `hw -r pci`; I don't know
why. ;-}

# hw -r cpu

Report about cpu for unix.testdom.com on Fri Apr 1 09:15:41 2011

There are 4 CPUs on this system.

hangs after four screens running worms see below.

hw -r mp

displays several screens then hangs. Piping to less
allows paging up and down but terminal session hangs
on quit from less.

after q and ctl-C,delete,ctl-\ no immediate response
but checking back later I see:

source_bus: 0x00 PCI
source_irq: 0x6a PCI INT_C#



q

/bin/ksh: 3622 Quit(coredump)
#
and the telnet session is once again working.

Anyone have a CPU benchmark program that is not dependent upon
disk subsystem speed?

Zillions all over the place.

I often use:

time awk 'BEGIN{for(i=0;i<10000000;i++);}'

which happens to take about 1.00s on this 2004 Xeon Nocona 3.6GHz
(NetBurst 90nm) system.

Bela<



One console screen running worms -R:

# hw cpu | grep performs
CPU 1 performs like a 1478Mhz Intel Pentium greater than III
CPU 2 performs like a 1478Mhz Intel Pentium greater than III
CPU 3 performs like a 1478Mhz Intel Pentium greater than III
CPU 4 performs like a 1478Mhz Intel Pentium greater than III

# time awk 'BEGIN{for(i=0;i<10000000;i++);}'

real 0m0.50s
user 0m0.50s
sys 0m0.00s
#

Two console screens running worms -R:
# hw cpu | grep performs
CPU 1 performs like a 1478Mhz Intel Pentium greater than III
CPU 2 performs like a 1478Mhz Intel Pentium greater than III
CPU 3 performs like a 1478Mhz Intel Pentium greater than III
CPU 4 performs like a 1478Mhz Intel Pentium greater than III
*Slow to respond to hw cpu* displayed after noticeable delay

# time awk 'BEGIN{for(i=0;i<10000000;i++);}'

real 0m0.50s
user 0m0.50s
sys 0m0.00s
#

Three screens running worms -R

# hw cpu | grep performs
CPU 1 performs like a 1478Mhz Intel Pentium greater than III
CPU 2 performs like a 1478Mhz Intel Pentium greater than III
CPU 3 performs like a 1478Mhz Intel Pentium greater than III
CPU 4 performs like a 1478Mhz Intel Pentium greater than III

No delay displayed with no noticeable delay

# time awk 'BEGIN{for(i=0;i<10000000;i++);}'

real 0m0.56s
user 0m0.56s
sys 0m0.00s
#

Four screens running worms -R

# hw cpu | grep performs
*never responds* telnet session locked up, can't kill it
ctrl-Z, ctrl-C, delete, ctrl-\ all ignored.

Logged in on new telnet session:

# time awk 'BEGIN{for(i=0;i<10000000;i++);}'

real 0m0.66s
user 0m0.66s
sys 0m0.00s
#

*Five console screens screens running worms -R*
10:13am up 9:31, 7 users, load average: 5.80, 5.16, 3.09
User Tty Login@ Idle JCPU PCPU What
root tty04 9:47am 25 10:54 10:54 /usr/games/worms -R
root tty06 9:56am 16 1:37 1:37 /usr/games/worms -R
root tty05 9:48am 18 9:33 9:29 /usr/games/worms -R
root tty08 10:04am 8 1:27 1:27 /usr/games/worms -R
root tty07 9:59am 13 6:53 6:53 /usr/games/worms -R
root ttyp0 8:44am 12 3 - hw cpu
root ttyp2 10:09am - 2 - w


| |
| system | user status sys cs int tr |
| ------------- ------ --- -- --- -- |
| +--------------------+ |
| CPU1 |||||||||||||||||||| | ACTIVE 237777 106 1 |
| +--------------------+ |
| +--------------------+ |
| CPU2 | | ACTIVE 82 14 112 0 |
| +--------------------+ |
| +--------------------+ |
| CPU3 | | ACTIVE22288524276 13431 1 |
| +--------------------+ |
| +--------------------+ |
| CPU4 ||||| | ACTIVE118399449655 24563 0 |
| +--------------------+ |
| |
|

# time awk 'BEGIN{for(i=0;i<10000000;i++);}'

real 0m0.54s
user 0m0.54s
sys 0m0.00s
# time awk 'BEGIN{for(i=0;i<10000000;i++);}'

real 0m0.71s
user 0m0.70s
sys 0m0.00s
# time awk 'BEGIN{for(i=0;i<10000000;i++);}'

real 0m0.72s
user 0m0.72s
sys 0m0.00s
# time awk 'BEGIN{for(i=0;i<10000000;i++);}'

real 0m0.53s
user 0m0.53s
sys 0m0.00s
#

Telnet session sluggish. Noticeable delay in typing and
seeing typed character appear on screen.

Five console screens running worms makes system effectively
unusable as keystrokes response in telnet sessions delayed
past annoying.

Might be a tunable parameter as programs like custom -> verify
software -> quick seem to run ok once started.

Woo!
# time cat /etc/ptest
1 1 2 3 4 5 6 7 8 910 1 2 3 4 5 6 7 8 920 1 2 3 4 5 6 7 8 930 1 2 3 4 5
2 1 2 3 4 5 6 7 8 910 1 2 3 4 5 6 7 8 920 1 2 3 4 5 6 7 8 930 1 2 3 4 5
3 1 2 3 4 5 6 7 8 910 1 2 3 4 5 6 7 8 920 1 2 3 4 5 6 7 8 930 1 2 3 4 5
4 1 2 3 4 5 6 7 8 910 1 2 3 4 5 6 7 8 920 1 2 3 4 5 6 7 8 930 1 2 3 4 5
....
766 1 2 3 4 5 6 7 8 910 1 2 3 4 5 6 7 8 920 1 2 3 4 5 6 7 8 930 1 2 3 4 5
767 1 2 3 4 5 6 7 8 910 1 2 3 4 5 6 7 8 920 1 2 3 4 5 6 7 8 930 1 2 3 4 5
768 1 2 3 4 5 6 7 8 910 1 2 3 4 5 6 7 8 920 1 2 3 4 5 6 7 8 930 1 2 3 4 5

real 0m13.04s
user 0m0.00s
sys 0m0.00s
#
Normally runs quickly but this run has many pauses in the display and real time
is appropriate for the delays observed.

With custom running software -> verify system -> quick

| |
| system | user status sys cs int tr |
| ------------- ------ --- -- --- -- |
| +--------------------+ |
| CPU1 |||||||||||||||||||| | ACTIVE 13528861 768 9 |
| +--------------------+ |
| +--------------------+ |
| CPU2 ||||||| | ACTIVE16544414613 7894 815 |
| +--------------------+ |
| +--------------------+ |
| CPU3 |||||||||||||| | ACTIVE39668721124 10547 1045 |
| +--------------------+ |
| +--------------------+ |
| CPU4 |||||||||| | ACTIVE54026327742 14083 940 |
| +--------------------+ |
| |
| |
|

time custom with no repair and mis keys trying to exit due
to delay in telnet keyboard response.

real 0m35.03s
user 0m16.11s
sys 0m3.78s
#

run on the console with fix and exit from custom shows:

real 0m29.73s
user 0m16.43s
sys 0m3.53s

There is no delay with keystrokes on the console screens
with five instances of worms running.

10:48am up 10:06, 12 users, load average: 3.77, 4.00, 4.58
User Tty Login@ Idle JCPU PCPU What
root tty04 9:47am 1:01 19:00 19:00 /usr/games/worms -R
root tty02 10:43am 4 21 - /bin/ksh
root tty06 9:56am 51 4:19 4:19 /usr/games/worms -R
root tty05 9:48am 54 18:01 17:57 /usr/games/worms -R
root tty08 10:04am 43 8:14 8:14 /usr/games/worms -R
root tty07 9:59am 49 25:53 25:53 /usr/games/worms -R
root tty09 10:44am 3 - - /bin/ksh
root ttyp0 8:44am 11 3 - /bin/ksh
root ttyp1 10:30am 18 - - /bin/ksh
root ttyp2 10:09am 8 2 - /bin/ksh
root ttyp3 10:19am 21 - - /bin/ksh
root ttyp4 10:21am - 40 - w
#

After killing all instances of worms and running
time cat /etc/ptest on a telnet session (much better
then the 13 seconds with 5 worms running):

# time cat /etc/ptest
1 1 2 3 4 5 6 7 8 910 1 2 3 4 5 6 7 8 920 1 2 3 4 5 6 7 8 930 1 2 3 4 5
2 1 2 3 4 5 6 7 8 910 1 2 3 4 5 6 7 8 920 1 2 3 4 5 6 7 8 930 1 2 3 4 5
3 1 2 3 4 5 6 7 8 910 1 2 3 4 5 6 7 8 920 1 2 3 4 5 6 7 8 930 1 2 3 4 5
4 1 2 3 4 5 6 7 8 910 1 2 3 4 5 6 7 8 920 1 2 3 4 5 6 7 8 930 1 2 3 4 5
....
767 1 2 3 4 5 6 7 8 910 1 2 3 4 5 6 7 8 920 1 2 3 4 5 6 7 8 930 1 2 3 4 5
768 1 2 3 4 5 6 7 8 910 1 2 3 4 5 6 7 8 920 1 2 3 4 5 6 7 8 930 1 2 3 4 5

real 0m0.01s
user 0m0.00s
sys 0m0.00s
#

Network performance took a nose dive when the system was loaded
with five instances of worms. Everything run on other console
screens were not a problem. The NIC shares IRQ with the PERC 6/iR
controller. But worms does not hit the hard disk so that does
not explain the sluggish network performance.

device address vec dma comment
======== ============= === === ================================================
kernel - - - rel=3.2v5.0.7 kid=2003-02-18
cpu - - - unit=1 family=6 type=gt PentIII
cpuid - - - unit=1 vend=GenuineIntel tfms=0:6:14:5(0)
fpu - 13 - unit=1 type=80387-compatible
pci 0xcf8-0xcff - - am=1 sc=0 buses=256
PnP - - - nodes=0
clock - - - type=TSC/2.659983697Ghz
serial 0x3f8-0x3ff 4 - unit=0 type=Standard nports=1 base=0 16550A/16
serial 0x2f8-0x2ff 3 - unit=1 type=Standard nports=1 base=8 16550A/16
console - - - unit=vga type=0 num=12 scoansi=1 scroll=50
floppy 0x3f2-0x3f7 6 2 unit=0 type=135ds18
udi - - - UDI environment
adapter - - - ha=0 type=usb_msto UDI SCSI HBA
adapter 0xfc00-0xfc80 15 - type=lsil ha=0 id=112 Chip=1068E 10409
adapter 0xecb0-0xecb7 10 - type=IDE ctlr=1 dvr=wd
bnxii0 - 15 - chip=5716C mem=DA000000 addr=78:2b:cb:0b:78:d4
cd-rom - - - type=IDE ctlr=1 cfg=mst unit=0 dvr=Srom->wd
disk - - - type=S ha=0 id=0 lun=0 bus=0 ht=lsil unit=0
Sdsk - - - cyls=17784 hds=255 secs=63 unit=0 fts=stdb
Sdsk-0 - - - Vnd=Dell Prd=VIRTUAL DISK Rev=1028
cpu - 255 - unit=2 family=6 type=gt PentIII
cpuid - - - unit=2 vend=GenuineIntel tfms=0:6:14:5(0)

# netstat -m
streams allocation:
config alloc free total max fail
stream 6144 95 6049 1397 98 0
queues 452 197 255 2803 203 0
mblks 612 353 259 197699 527 0
buffer headers 826 706 120 7711 731 0
class 1, 64 bytes 192 4 188 101383 131 0
class 2, 128 bytes 64 0 64 29289 37 0
class 3, 256 bytes 48 8 40 26112 35 0
class 4, 512 bytes 24 6 18 1178 22 0
class 5, 1024 bytes 24 0 24 1373 24 0
class 6, 2048 bytes 286 269 17 1743 286 0
class 7, 4096 bytes 0 0 0 0 0 0
class 8, 8192 bytes 1 0 1 16 1 0
class 9, 16384 bytes 1 0 1 8 1 0
class 10, 32768 bytes 0 0 0 0 0 0
class 11, 65536 bytes 0 0 0 0 0 0
class 12, 131072 bytes 0 0 0 0 0 0
class 13, 262144 bytes 0 0 0 0 0 0
class 14, 524288 bytes 0 0 0 0 0 0
total configured streams memory: 25024.00KB
streams memory in use: 581.23KB
maximum streams memory used: 695.59KB

The T310 BIOS assignment screen shows

Video IRQ 6
NIC1 15
NIC2 5 -> 11
Slot1 PERC6/iR 15
South Bridge device 16 15
USB EHCI controller 2 14
USB EHCI controller 1 14
Sata 1 11 -> 14
Sata 2 10

The items shown at 15 all change simultaneously they can't
be separated.

I'll reconfigure the UNIX kernel to use the other on-board
NIC to move the NIC to IRQ11. I'll move SATA1 to 14 as SATA2
is supporting the DVD-drive.

======== ============= === === ================================================
kernel - - - rel=3.2v5.0.7 kid=2003-02-18
cpu - - - unit=1 family=6 type=gt PentIII
cpuid - - - unit=1 vend=GenuineIntel tfms=0:6:14:5(0)
fpu - 13 - unit=1 type=80387-compatible
pci 0xcf8-0xcff - - am=1 sc=0 buses=256
PnP - - - nodes=0
clock - - - type=TSC/2.659983947Ghz
serial 0x3f8-0x3ff 4 - unit=0 type=Standard nports=1 base=0 16550A/16
serial 0x2f8-0x2ff 3 - unit=1 type=Standard nports=1 base=8 16550A/16
console - - - unit=vga type=0 num=12 scoansi=1 scroll=50
floppy 0x3f2-0x3f7 6 2 unit=0 type=135ds18
udi - - - UDI environment
adapter - - - ha=0 type=usb_msto UDI SCSI HBA
adapter 0xfc00-0xfc80 11 - type=lsil ha=0 id=112 Chip=1068E 10409
adapter 0xecb0-0xecb7 15 - type=IDE ctlr=1 dvr=wd
bnxii0 - 10 - chip=5716C mem=DC000000 addr=78:2b:cb:0b:78:d5
cd-rom - - - type=IDE ctlr=1 cfg=mst unit=0 dvr=Srom->wd
disk - - - type=S ha=0 id=0 lun=0 bus=0 ht=lsil unit=0
Sdsk - - - cyls=17784 hds=255 secs=63 unit=0 fts=stdb
Sdsk-0 - - - Vnd=Dell Prd=VIRTUAL DISK Rev=1028
cpu - 255 - unit=2 family=6 type=gt PentIII
cpuid - - - unit=2 vend=GenuineIntel tfms=0:6:14:5(0)

Well, moving the IRQ's around in BIOS did not resolve the
issue with network slowdown with five console terminals
running worms -R. I need to wrap up this installation
and deliver the system to the customer. We will see if
they have any problems.

After I write the above, I went back and ran time cat /etc/ptest
again (the first time it came in at 13 seconds) and see:

# time cat /etc/ptest
1 1 2 3 4 5 6 7 8 910 1 2 3 4 5 6 7 8 920 1 2 3 4 5 6 7 8 930 1 2 3 4 5
2 1 2 3 4 5 6 7 8 910 1 2 3 4 5 6 7 8 920 1 2 3 4 5 6 7 8 930 1 2 3 4 5
3 1 2 3 4 5 6 7 8 910 1 2 3 4 5 6 7 8 920 1 2 3 4 5 6 7 8 930 1 2 3 4 5
4 1 2 3 4 5 6 7 8 910 1 2 3 4 5 6 7 8 920 1 2 3 4 5 6 7 8 930 1 2 3 4 5
....
766 1 2 3 4 5 6 7 8 910 1 2 3 4 5 6 7 8 920 1 2 3 4 5 6 7 8 930 1 2 3 4 5
767 1 2 3 4 5 6 7 8 910 1 2 3 4 5 6 7 8 920 1 2 3 4 5 6 7 8 930 1 2 3 4 5
768 1 2 3 4 5 6 7 8 910 1 2 3 4 5 6 7 8 920 1 2 3 4 5 6 7 8 930 1 2 3 4 5

real 0m0.18s
user 0m0.00s
sys 0m0.00s

and custom run on a telnet session:

# time custom
real 0m25.74s
user 0m14.39s
sys 0m3.12s
#

Keystrokes on the telnet session are quick with no precipitable delay.

# w
12:15pm up 20 mins, 12 users, load average: 5.08, 5.10, 3.52
User Tty Login@ Idle JCPU PCPU What
root tty07 11:59am 15 4:26 4:26 /usr/games/worms -R
root tty08 11:59am 15 3:57 3:57 /usr/games/worms -R
root tty04 11:57am 16 1:03 1:03 /usr/games/worms -R
root tty06 11:58am 15 2:10 2:10 /usr/games/worms -R
root tty05 11:58am 16 3:44 3:44 /usr/games/worms -R
root tty09 11:59am 15 2:05 2:05 /usr/games/worms -R
root tty02 11:59am 16 - - mpstat
root ttyp0 12:01pm 7 - - /bin/ksh
root ttyp1 12:02pm 10 9 2 /usr/lib/edge/bin/edge.tape -arg

root ttyp2 12:05pm 10 - - mpstat
root ttyp3 12:06pm 7 36 - w
root ttyp4 12:08pm - 18 - /bin/ksh
#

So moving the BIOS IRQ assignments around did change the network
performance. I note when viewing the console screens running
worms that they all are moving without any delays or hesitations
but they all seem to be throttled from what I observed in the
screens before changing the BIOS assignments.

Whoa! After edge completed the listing of last night's backup I see

rw------- 0/0 146 Mar 29 12:22 2011 ./boot, 1 blocks (compressed)
rw------- 0/0 73 Mar 29 12:23 2011 ./stty.temp, 1 blocks
rwxr-xr-x 0/2 0 Apr 1 01:07 2011 ., 0 blocks (Directory)
rw-r--r-- 0/3 2113 Nov 27 23:11 2002 /etc/ptest, 5 blocks (compressed)
Exit Status = 0

766 1 2 3 4 5 6 7 8 910 1 2 3 4 5 6 7 8 920 1 2 3 4 5 6 7 8 930 1 2 3 4 5
767 1 2 3 4 5 6 7 8 910 1 2 3 4 5 6 7 8 920 1 2 3 4 5 6 7 8 930 1 2 3 4 5
768 1 2 3 4 5 6 7 8 910 1 2 3 4 5 6 7 8 920 1 2 3 4 5 6 7 8 930 1 2 3 4 5

real 0m13.00s
user 0m0.00s
sys 0m0.00s
#

And keystrokes again are delayed.

Restart edge listing function and performance on the network is again good:

766 1 2 3 4 5 6 7 8 910 1 2 3 4 5 6 7 8 920 1 2 3 4 5 6 7 8 930 1 2 3 4 5
767 1 2 3 4 5 6 7 8 910 1 2 3 4 5 6 7 8 920 1 2 3 4 5 6 7 8 930 1 2 3 4 5
768 1 2 3 4 5 6 7 8 910 1 2 3 4 5 6 7 8 920 1 2 3 4 5 6 7 8 930 1 2 3 4 5

real 0m0.17s
user 0m0.00s
sys 0m0.00s
#
With the edge listing running mpstat shows:

| |
| system | user status sys cs int tr |
| ------------- ------ --- -- --- -- |
| +--------------------+ |
| CPU1 |||||||||||||||||||| | ACTIVE 161 5970 505 1 |
| +--------------------+ |
| +--------------------+ |
| CPU2 | | ACTIVE 92 170 315 0 |
| +--------------------+ |
| +--------------------+ |
| CPU3 | | ACTIVE 8205 2259 1369 1 |
| +--------------------+ |
| +--------------------+ |
| CPU4 | | ACTIVE188339 9790 4977 1 |
| +--------------------+ |
| |
|

Stopping Edge listing of the backup media:

| |
| system | user status sys cs int tr |
| ------------- ------ --- -- --- -- |
| +--------------------+ |
| CPU1 |||||||||||||||||||| | ACTIVE 227170 106 1 |
| +--------------------+ |
| +--------------------+ |
| CPU2 | | ACTIVE 39 10 106 1 |
| +--------------------+ |
| +--------------------+ |
| CPU3 | | ACTIVE 39630 9310 4787 1 |
| +--------------------+ |
| +--------------------+ |
| CPU4 || | ACTIVE89298645029 22585 0 |
| +--------------------+ |
|


This is funky shit. The telnet slowdown my have nothing to do with
IRQ assignments. Maybe the CPU is clocking up under load? No way to
tell as hw locks up.

--
Steve Fabac
S.M. Fabac & Associates
816/765-1670
.

Relevant Pages