WPC  
      2      B       P   Z            Courier 10cpi 3| x              px 6X   @ 8	 X@DOS Text Printer                     DOTEXPRI.PRS x 
   @           X@3| x           2      B         V      Z   .   "  Sh
^CCd~~ȘCCC~CCCC~~~~~~~~~~CChȔ~so8okw~ȐzCCCddCkskskJss00k0ssssFdJsooo`YdY~~~  ~~~   C ~~~~~~ d~z0 kkkkkĪksksksksk80808080sssssssssok~ssoswsk~~  ~~~~~~~~~~~~  s~~~~~~~0~~~~8z`~~~  ~~~~kIk0   ~~~~ȵ  ~~  d~~~~~~~~s~~~  sȮ~  z`C     dd   C      CC/Nd ~~Choo~~~QQ~NNdd~N CC dd~JJ~~zzddd zz     Cd      d     N"Ȑdj             d  dCCȐwȐ Cd~o~ȐȔdC~d~sȐkȐ ȐȧzȐ       Uw                                                                                                 ŐdȐ      Y   Y                           C   C   C   C                                                               Ŕo~k  odks]zUsk80dhoo~Usswkww Y~]k`sC~ko   CC~~~ ~                                                                  ~~~                                           C  sdYC  Courier 10cpi    ? x x x 
     ^x 6X   @ 8	 X@DOS Text Printer                     DOTEXPRI.PRS Xy    @ ;         X@ 2                                            XX  XX    ***************************************************************************
AXS Benchmarks`!%H 02.28.94
***************************************************************************

Technical Information:
 X
 Servers: `	`	  X Gateway 486/66e, 32MB, Intel EtherExpress 16 10BaseT
NIC, Novell 3.12, 32Bit Eisa SCSI Disk Controller, 1Gig
Driveƀ%

 X
 Workstations:  Gateway 486, 8MB, 16bit NIC.  All configured with Qemm
6.02, MSDos 5.0, IPX v3.10, Netx v3.21ƀ%
 X
  X  X`	`	  X ƀ%
 X
 AXS NLM Settings: `	 ©I160 D75 U50 W500.  All other settings at
default level.ƀ%

 X
 Program: `	`	  X Clipper 5.01a.  Blinker 3.0 (Real Mode).  AXS version
2.01ƀ%

Database Information:  All databases have 5 indexes, ranging from single
to multiple field keys.  Each database also has at
least 2 indexes with numeric keys.ƀ%


    X`	`	 Name  hh# #Records- pp2 #Fields  < xxA RecSizeƀ%`	

 1  `	`	 SALES.DBF hh# 6125( - pp2 207   < xxA 231

 2  `	`	 CLIENTS.DBFhh# 9924( - pp2 227   < xxA 301

 3  `	`	 TRANS.DBF hh# 12246- pp2 497   < xxA 355

 4  `	`	 ITEMS.DBF hh# 5652( - pp2 267   < xxA 190

Tests:

 X
 1. X Skip through each of the 4 Databases in a "Do While !Eof()" loop. 
All indexes open, active Order set to Index # 1.ƀ%

 X
 2. X Skipping with Relations:ƀ%
 X
  X  X`	`	 Set Relation to Sales>CliNum into Clientsƀ%`	

  `	`	 Set Relation to Sales>SaleNum Into Trans

  `	`	 Set Relation to Trans>PackCode Into Items


   Do a Go Top on Sales and skip through the entire database in a
"Do While !Eof()" loop.ƀ%

 X
 3. X 12571 Random Seeks in each of the 4 databases.ƀ%

 X
  X I created a database that has two fields: "Key" and "Alias". 
Spin through the database and (Alias)>(dbSeek(Key)).  I
specifically made this database to contain completely random key
values, some may not even be found.ƀ%

 X
 4. X Append 500 Records.  Append 500 records to each of the 4
databases, updating the ALL the Key fields with some random
value.  Essentially 2000 records added and 2000 index keys added.ƀ%

 X
 5. X Update Records.   Change the All the Key fields in 500 Random
records throughout each of the 4 databases.  Essentially 2000
index key updates.ƀ%

 X
 Note: `	`	 The Benchmark program was run 3 times in a row for each test
phase (Single, Two User, Eight user) to get an average of
the values.ƀ%`	
   .         ...XX  ԌTest Results (all numbers in seconds):

 =======================================================
," Single User Test
   X`	hp x (#%'0*,.8135@8:<H?A                      p                    X    p                      h                     XX 
 Testhh# DBFNTX Driver  < NTXAX Driver
  1 `	hh# 102  < 107ƀ%h
 2hh# 20  < 20
 3hh# 62  < 59
 4hh# 97  < 77
 5hh# 134  < 77

 =======================================================
$ Two User Test

 Testhh# DBFNTX Driver  < NTXAX Driver
  1hh# 135  < 131ƀ%
 2hh# 27  < 25
 3hh# 65  < 70
 4hh# 211  < 125
 5hh# 218  < 98

 =======================================================
h# Eight User Test

  Testhh# DBFNTX Driver  < NTXAX Driverƀ%
  1hh# 620  < 372ƀ%
 2hh# 132  < 72
 3hh# 161  < 201
 4hh# 380  < 580
 5hh# 330  < 301
 =======================================================

   h                       `	hp                    XX 
Packet test:

 X
 During the eight user test I tried to measure the number of packets
transmitted on the LAN.  I'm not really sure about the meaning of
these numbers.  I felt that by getting these measurements I would be
able to determine the reduction in network traffic by using AXS, but
I'm not completely convinced that it's an accurate measurement.ƀ%


   DBFNTX Driver: `	hh#  307500ƀ%h

  NTXAX Driver: hh#  97500ƀ%h

Other notes:

 X
 I did watch the Server Utilization during all of the test phases. 
Basically, the Server Utilization was higher when using the AXS
driver.  This was expected.  However, I should note that during the
AXS eight user test, the Utilization hovered between 85% and 99%,
while on the NTX test it barely made it over 75%.  Again, I am unsure
on the exact affect this high utilization has on the overall
performance of the server.  There are a lot of variables that should
be considered when evaluating these numbers.ƀ%

 X
 I did not include the datafiles with this test information because of
the large size of the ZIP file (over 2MB).  And also because I pulled 
the datafiles from one of the systems they had running at KT, so it's
real data  real messy data too!ƀ%   .        ...XX  Ԍ X
 I did not test with the CDX driver because I was running into problems
with the DBFCDX RDD in Clipper 5.2c.  I would like to run these
benchmarks under the AXSCDX driver for a more complete comparison. 
Soon, I hope.ƀ%

 X
 I'm not passing any judgment on AXS, this is just the results of my
tests.  If required, based on feedback, I'll reconfigure the benchmark
program/data and rerun the tests.  I want to be sure that AXS will
provide a worthwhile speed increase for "our" application in "our"
environment before I recommend it.ƀ%

 X
 Send any info to Mike Smith @ 71034,1653.