| Here is the notice of the new system characterization guides out of
Europe's Systems Engineering Group. I think you'll find these
documents have a lot of valuable information in them, especially for
anyone who is setting up an LAVc.
Enjoy,
Jim
<<< BYENG0::SYSTEM_2:[NOTES$LIBRARY]SEE_MCAD_SYSTEMS.NOTE;1 >>>
-< ESG MCAD Solution Systems Engineering, Munich >-
================================================================================
Note 15.0 EUCLID V2.1 System and Char. Guides No replies
BYENG0::RHUBERT "Richard, ESG Sys. Eng., Europe" 437 lines 17-Jul-89 11:27
--------------------------------------------------------------------------------
+---+---+---+---+---+---+---+
| d | i | g | i | t | a | l | I N T E R O F F I C E M E M O
+---+---+---+---+---+---+---+
| Engineering Systems Group | From : RICHARD HUBERT
+---------------------------+ Date : 14-Jul-1989
| International Engineering | Dept.: Eng. Syst. Group, Europe
+---------------------------+ Loc. : RTO
Tel. : [49](89)9591-1075
DTN : 865-1075
Mail : Byeng0{ByengE}::RHUBERT
To: see distribution list
Subject: AVAILABILITY OF EUCLID V2.1 MCAD SYSTEM PERFORMANCE and
CHARACTERIZATION GUIDES
The EUCLID V2.1 (VMS 5.1 VAXcluster platforms) System Performance and
Characterization Guides are now available from ESGSEE Munich.
There are two versions of the guide:
- one for DIGITAL Internal Use Only and
- one intended for external distribution if desired.
THESE GUIDES HAVE BEEN SUBMITTED TO DCC MARKETING (MUNICH)
FOR FINAL APPROVAL AND FORMAL DISTRIBUTION.
The table of contents of the "Internal" version, which is the most complete, is
appended to the bottom of this mail. In short, the guide contains sections on
- Why, how, what was done.
- Conclusions and discussion in summary form. This also includes
configuration guidelines.
- MCAD workloads study and performance details.
- Tuning
- Modeling (Predictive Modeling). The progress that was made and
conclusions drawn using modeling in the short time span between
the availability of all the lab results and the publication of
the guide.
- Some EUCLID IS installation tips.
The documents are available in LN03 format ONLY due to the sixel graphics
content. They may be copied from: (Node Byeng0 has the number 49.750)
Byeng0::see$disk:[esgsee.netcopy]xyz.ln03 where "xyz" is
- Euclid_v21_char_guide_intern (This is the DIGITAL internal version)
- Euclid_v21_char_guide_open (This is the non-internal version)
If you are not interested in the results and conclusions and just want to
see what we suggest for tuning or what was done with modeling or how we
designed and characterized the MCAD Workloads, then the sections for these
are available in the following (LN03) files. NOTE: All references made from
these sections to other sections in the document are missing since they were
not compiled together.These are subsets of the "DIGITAL internal" document
given above. They may be copied from: (Node Byeng0 has the number 49.750)
Byeng0::see$disk:[esgsee.netcopy]xyz.ln03 where "xyz" is
- Appendix_workloads (MCAD Workload design and profiles)
- Appendix_tuning (Euclid IS tuning information)
- Appendix_modeling (Predictive modeling results/progress)
The table of contents of the "DIGITAL internal" version (the most complete)
follows.
Regards,
Richard
-----------------------------------TABLE OF CONTENTS-------------------------
MATRA DATAVISION EUCLID IS V2.1
System and Characterization Guide
July 14 1989
This System and Characterization Guide provides performance,
tuning and configuration guidelines for VAXcluster and single
user systems in an MCAD EUCLID IS environment.
THIS DOCUMENT IS FOR INTERNAL USE ONLY.
It contains sensitive performance and operation data con-
cerning the tested configurations. A modified version is
available which summarizes the important facts but omits
information which may be subject to misinterpretation,
only the modified version should be distributed to persons
outside DIGITAL.
Revision/Update Information: 14-July-89 ESGSEE (Munich)
Digital Equipment Corporation
CONTENTS
Preface............................................... viii
Chapter 1 INTRODUCTION................................... 1
1.1 MDTV EUCLID V2.1................................... 1
1.2 Hardware Base...................................... 1
1.2.1 Single User System.............................. 2
1.2.2 VAXcluster Systems.............................. 2
1.2.2.1 CI-based VAXcluster Configurations........... 4
1.2.2.2 Local Area VAXcluster Configurations......... 4
1.2.2.3 Mixed-Interconnect VAXcluster Configurations. 5
1.2.3 Two Sample LAVc Systems......................... 5
1.2.3.1 LAVc with Server/Workstation................. 5
1.2.3.2 LAVc with Dedicated Server................... 7
1.3 Software Structure................................. 11
1.3.1 MDTV EUCLID V2.1................................ 11
1.3.2 VAX FORTRAN..................................... 11
1.3.3 Component Interfaces............................ 11
Chapter 2 HOW HAS CHARACTERIZATION BEEN PERFORMED........ 13
2.1 Questions to Answer................................ 13
2.2 Difficulties With Complex Applications............. 13
2.3 Steps of the Characterization Process.............. 15
2.3.1 Step 1: Definition.............................. 15
2.3.1.1 Definition of Configurations................. 16
2.3.1.2 Definition of Workloads...................... 16
2.3.1.3 Use of the Workloads......................... 19
iii
2.3.2 Step 2: Test Execution.......................... 20
2.3.2.1 Test Automation.............................. 20
2.3.2.2 Monitoring Tools............................. 20
2.3.2.3 Testing One Particular Configuration......... 21
2.3.3 Step 3: Data Reduction and Evaluation of Results 22
2.3.3.1 Comparison of Interactive Performance........ 22
2.3.3.1.1 Elapsed Time (ET)......................... 23
2.3.3.1.2 Computed Elapsed Time (CET)............... 24
2.3.3.1.3 Computed Elapsed Time Factor (CETF) and
Reference Configuration................... 24
2.3.3.2 Resource Utilization......................... 26
2.3.3.2.1 CPU-Utilization, %-busy................... 26
2.3.3.2.2 Memory Utilization, %-used................ 27
2.3.3.2.3 Disk Storage Utilization, %-used.......... 27
2.3.3.2.4 Network Adapter Utilization, %-used....... 28
2.3.4 Step 4: Generalization and Applicability of
Results......................................... 29
2.3.4.1 Extrapolation of Results..................... 29
2.3.4.2 Predictive Modeling.......................... 30
Chapter 3 WHAT HAS BEEN CHARACTERIZED.................... 31
3.1 Tested Configurations.............................. 31
3.2 Configurations Which Have Not Been Tested.......... 32
3.2.1 Diskless Workstations........................... 32
3.2.2 8 MByte Memory Workstations..................... 32
3.2.3 VAXstation 3200 Family.......................... 32
3.2.4 VAXstation 3520/3540 Family..................... 33
3.3 List of Tested Configurations...................... 33
iv
Chapter 4 BASIC RESULTS AND SUGGESTIONS.................. 39
4.1 Comparison of Workstation Types.................... 39
4.1.1 Obligatory Hardware Components.................. 39
4.1.1.1 VAXstation 3100.............................. 40
4.1.1.2 VAXstation 3500.............................. 41
4.1.1.3 VAXstation 3200.............................. 42
4.1.1.4 VAXstation II/GPX............................ 42
4.1.1.5 Summary of Functional Comparison of
Workstation Types............................ 43
4.1.2 Performance of Workstation Types................ 45
4.1.2.1 VAXstation 3100 vs. VAXstation 3500.......... 45
4.1.2.2 VAXstation 3xxx vs. VAXstation II/GPX........ 49
4.1.2.3 Influence of Physical Memory on Performance.. 51
4.1.3 Summary of Workstation Comparison............... 52
4.2 Local Area VAXcluster Configurations............... 54
4.2.1 Functional Aspects.............................. 54
4.2.1.1 LAVc Workstations vs. Networked Single User
Systems...................................... 54
4.2.1.2 Server/Workstation vs. Dedicated Server...... 56
4.2.2 Performance Aspects............................. 58
4.2.2.1 LAVc with Server/Workstation VAXstation 3100. 60
4.2.2.2 LAVc with Dedicated VAXserver 3600........... 64
4.2.2.3 Display-Server Configuration for VAXstation
II and VAXstation 2000 series................ 66
4.2.3 Summary of LAVc Configurations.................. 67
Chapter 5 DETAILED RESULTS AND CONCLUSIONS............... 71
5.1 A Method to Estimate Hardware Resource Needs and
Bottlenecks........................................ 71
5.2 RMS and File Characteristics in the EUCLID
Environment........................................ 73
v
5.3 Significant Changes in VMS 5.0 that Benefit
MCAD/EUCLID Performance in a VAXcluster............ 74
5.3.1 Lock Database Rebuilds During Cluster State
Transitions..................................... 74
5.3.2 AUTOGEN with FEEDBACK........................... 75
5.3.3 MSCP Server..................................... 75
5.3.4 Backup.......................................... 76
5.3.5 RMS............................................. 76
Appendix A DESCRIPTION OF WORKLOADS...................... 77
A.1 General Information................................ 77
A.2 Workloads of the Design-Engineering Group.......... 81
A.2.1 2D-Monoview..................................... 81
A.2.2 2D-Multiview.................................... 82
A.2.3 Detailed 2D/3D Drafting......................... 83
A.2.4 Parametric Drafting............................. 84
A.2.5 Retrieve and Re-Construct....................... 90
A.2.6 3D-Surface Modelling............................ 91
A.2.7 3D-Solid Modelling 1 (Mouse).................... 92
A.2.8 3D-Solid Modelling 2 (Backplane)................ 93
A.2.9 Bill of Materials............................... 94
A.3 Workloads of the Manufacturing-Engineering Group... 95
A.3.1 Mechanics / Kinematics (Robot).................. 95
A.3.2 NC-Milling...................................... 96
A.3.3 NC-Turning...................................... 98
A.4 Special Application Workload....................... 99
Appendix B CONDENSED DATA ON TESTED CONFIGURATIONS....... 101
B.1 Example of the Test Data Sets...................... 101
B.2 Dictionary of Test Cases........................... 104
vi
B.3 Selected Data...................................... 120
B.3.1 Workstation Comparison.......................... 120
B.3.2 LAVc with Dedicated VAXserver 3600.............. 123
B.3.2.1 Server VAXserver 3600........................ 123
B.3.2.2 Workstation VAXstation 3100.................. 125
B.3.2.3 Workstation VAXstation 3500.................. 127
B.3.3 LAVc with Server/Workstation VAXstation 3100.... 128
B.3.3.1 Server/Workstation VAXstation 3100........... 128
B.3.3.2 Workstation VAXstation 3100.................. 130
B.3.4 Display-Server Configuration for VAXstation
II/GPX.......................................... 132
B.3.4.1 Server VAXserver 3600........................ 132
B.3.4.2 Workstation VAXstation II/GPX................ 133
Appendix C INSTALLATION TIPS AND RECOMMENDATIONS......... 135
C.1 Setting Up a Euclid Display Server Configuration... 135
C.2 Setting Up an LN03R or LPS40 as an EUCLID Plotter.. 137
Appendix D TUNING RECOMMENDATIONS AND STRATEGIES......... 141
D.1 VAXserver 3600 as LAVc VAXmanager/Server, General
VAXmanager/Server.................................. 141
D.1.1 VMS Parameter Recommendations for VAXserver..... 142
D.2 VS3100 and VS3500 as LAVc VAXmanager, LAVc
Workstation, Single User Station................... 146
D.2.1 VMS Parameter Recommendations for VAXstations as
WS and VAXmanager............................... 146
D.3 EUCLID Display-Server Configuration................ 151
D.4 User Authorization Quotas for EUCLID Users......... 158
D.5 General and Miscellaneous.......................... 161
vii
Appendix E MODELING (A BEGINNING)........................ 163
E.1 The Models: Usage and Current Possibilities........ 163
E.1.1 VAX Performance Advisor and VME................. 165
E.1.2 Data Acquisition and Building LAVc Base Models.. 165
E.1.3 Validation of Model and Model Detail of Measured
Workloads....................................... 165
E.2 Current Work and Future Directions................. 182
E.2.1 LAVc : Modeling Critical Configuration Load and
Configuration Size Predictions.................. 182
E.2.2 Accurate Extension of LAVc Models............... 188
E.2.3 Work Planned and Tools Under Development........ 191
Appendix F REFERENCES.................................... 193
GLOSSARY.................................................... 195
FIGURES
1 Single User System Hardware Configuration......... 3
2 Small LAVc Hardware Configuration................. 5
3 Medium Sized LAVc Hardware Configuration.......... 9
4 Software Structure................................ 12
23 Resource Utilization Profile for 2D-Monoview...... 81
24 Hardcopy from 2D-Monoview Workload Result......... 81
25 Resource Utilization Profile for 2D-Multiview..... 82
26 Hardcopy from 2D-Multiview Workload Result........ 82
27 Resource Utilization Profile for Detailed
2D/3D-Drafting.................................... 83
28 Hardcopy from Detailed 2D/3D Drafting Workload
Result............................................ 83
29 Resource Utilization Profile for Parametric
Drafting.......................................... 84
viii
30 Resource Utilization Profile for Retrieve and
Re-Construct...................................... 90
31 Hardcopy from Retrieve and Re-Construct Workload
Result............................................ 90
32 Resource Utilization Profile for
3D-Surface-Modelling.............................. 91
33 Hardcopy from 3D-Surface Modelling Workload
Result............................................ 91
34 Resource Utilization Profile for
3D-Solid-Modelling 1.............................. 92
35 Hardcopy from 3D-Solid-Modelling Workload 1
Result............................................ 92
36 Resource Utilization Profile for
3D-Solid-Modelling 2.............................. 93
37 Hardcopy from 3D-Solid Modelling Workload 2
Result............................................ 93
38 Resource Utilization Profile for Bill of
Materials......................................... 94
39 Resource Utilization Profile for
Mechanics/Kinematics.............................. 95
40 Hardcopy from Mechanics/Kinematics Workload
Result............................................ 95
41 Resource Utilization Profile for NC-Milling....... 96
42 Hardcopy from Milling Workload Result............. 97
43 Resource Utilization Profile for NC-Turning....... 98
44 Hardcopy from Turning Workload Result............. 98
45 Resource Utilization Profile for Large
Assembly.......................................... 99
46 Hardcopy from Large Assembly Workload Result...... 99
ix
TABLES
1 Workloads......................................... 17
2 System Configurations for Workstation
Comparison........................................ 33
3 System Configurations for LAVc
Characterization.................................. 34
4 Features of Workstation Types..................... 44
5 Selection of LAVc Servers......................... 69
6 Workloads Used for Characterization............... 79
7 Example VMS Parameters for VAXmanager/server...... 142
8 VS3X00 Single User and LAVc Station, and
VS3100/3500 VAXmanager/server..................... 148
9 PQL SYSGEN Parameters for a Limited (8 MByte)
Memory Configuration.............................. 150
10 Example Page and Swapfile Sizes................... 151
11 Example VMS Parameters for server in
Display-Server Configuration...................... 152
12 Example VMS Parameters for display station in
Display-Server Configuration...................... 156
x
|