| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 680.1 | | BSS::JILSON | WFH in the Chemung River Valley | Wed Mar 12 1997 16:11 | 6 |
| Zip the .CPD file(s) and your .MEMORY file and post a pointer to it. Then
we'll see what we can find. I will be on vacation for a week starting on
Friday. If this si for a customer then IPMT it and make the same data
available in the IPMT case.
Jilly
|
| 680.2 | Data you suggested | MSE1::ENO | Holographic Memory | Wed Mar 12 1997 17:39 | 25 |
| File MSEDV1::BANK1.ZIP contains two memory files I have tried to
use and one days CPD files for the four systems.
File MSEDV1::BANK2.ZIP contains two memory files I have tried to
use and two additional days CPD files for the four systems.
I am checking on an IPMT case number, however, I believe the customer
could not build a model so they requested we do it. We accepted
since I never had a problem doing such plans in the past... This is
not just this one customer... I have additional customers with 8400s
who also cannot build models and have requested the same... I am very
concerned and puzzeled... I would rather have the customer use the
data collector with ECP but I am not the customer..
The systems are mirrored BRS sites with one AlphaServer 8400 Model
5/350 with 4 CPUs and a AlphaServer 2100A 5/300 with 2 CPUs at each
site connected VIA FDDI device type DEFPA. CI devices are CIPCA.
SCSI controllers are KZPSA HS devices are HSJ40 Firmware V27J-0.
All disks are RZ29B except for D635 and D1135 which are EZ31 and
$1$DKA0, $1$DKA100, $2$DKA0, $2$DKA100 which are RZ28D. The ethernet
controllers are DE435.
Let me known what you can find...
David
|
| 680.3 | | BSS::JILSON | WFH in the Chemung River Valley | Fri Mar 14 1997 10:05 | 33 |
| What I have so far.
This is a lot of data and you *will* need an Alpha to work on it to get any
kind of reasonable performance.
For BANK1.ZIP I was able to reduce all nodes except RLIV00. There appears
to be either problems with the CPD data or the customer is using PREFER to
change MSCP servers or REDUCE has a problem with the data. I suspect this
is the .CPD file you need to supply in the IPMT. I was using 10:00-11:00
I am still trying to use a few different time periods. With the size of
this data I would not suggest using time periods any longer than 1 hour and
30 minutes will probably be sufficient.
I haven't been able to get to BANK2.ZIP but this looks like the data is
even bigger.
I will have one last update today before going on vacation.
Jilly
Example of problems with RLIV00
REDUCE/BY_IMAGE/BEGIN=04-FEB:09:45/END=04-FEB:10:45 PSDC$RLIV00_1997FEB04.CPD
ECP-INFO, Driver for CPD ...
ECP-WARNING, Disk Server configuration has changed
.
.
could not find RLIV002$DUA1226
could not find RLIV002$DUA1226
ECP-WARNING, Disk configuration has changed
ECP-WARNING, Disk configuration has changed
ECP-WARNING, Image deletion event missing for process: 22c03a91
ECP-ERROR, Cannot allocate 34583520 bytes
|
| 680.4 | | BSS::JILSON | WFH in the Chemung River Valley | Fri Mar 14 1997 14:40 | 20 |
| Last update before I go on vacation for a week.
This data is large by any stretch of the imagination. I can't see how you
could work with it without an Alpha workstation with at least 128 meg. You
*need* large working set and virtual address space.
I was able to reduce 04-Feb 13:30-14:00 with little problems and I built a
model that looks reasonable from the ECP_ENO.MEMORY;1 list (although I
don't see the usefullness in your subclasses).
IMHO You need to spend a lot more time with the customer finding out
*EXACTLY* what they want to model then look at the data with PSPA and find
the best time period to model. I suspect that they/you are falling into
the trap of trying to create classes for every piece of work done on the
system instead of just a handful of classes needed to answer the question
at hand and are trying to model too long of a time period (ie > 1 hour). I
believe you can build a valid model out of this data but if you cannot
after trying again then you need to IPMT this.
Jilly
|
| 680.5 | A brief analysis was done but very brief... | MSE1::ENO | Holographic Memory | Tue Mar 25 1997 14:17 | 6 |
| Yes the data set is very large... And I discovered that the tool will loop if disk
configurations change which is the case for this customer... I use an Alpha 8400
with 2 GB memory and liberal working sets, but during the model build it still loops.
I have backed down to version 3.2, because version 4 kits simply crashes when I am
attempting to do any model modification...
|