| Oracle FAQ | Your Portal to the Oracle Knowledge Grid | |
 |
 | |||
| Oracle FAQ | Your Portal to the Oracle Knowledge Grid | |
|
| |||
Home -> Community -> Usenet -> c.d.o.server -> Re: NT vs Unix
Hello,
I feel that the responses have quickly degenerated along way from the
original question.
I have recently headed a validation team whose task was to qualify and kit
Oracle 805.
The platforms chosen were DIGITAL DU4.0d, SUN SOLARIS 2.6 and NT 4.0
Tip if you wish install 20 or 30 NT servers how would you do this.
For more than five installations its worth contacting Oracle and asking
for their
'Unofficial guide to Oracle Installer Command Line Silent installations' by
Mark Torstenson version 1.2
This works for both 7343 and 8050 on NT4.0
The machines for the park were standard configurations i.e.
DEC 4100 2 CPU 2 GB RAM and 200 GB disk striped 32k over 2 disks SUN 3500 2 CPU 2 GB RAM and one volume of 90 GB COMPAQ PROLIANT 4 CPU server 512 MB disk 40 GB disk only option permitted RAID5
We performed memory ,disk and general db tests using an identical load
simulator with the same 30 GB database in all configurations.
The aim was to deliver standard installable kits or packages. When you do
set up environments the only measure that you have exactly the same versions
of SQLNet, et al and associated libraries is a validation of the contents
of the UNIX.rgs NT.rgs in the install directory.
When and only when this is performed can we begin to bitch about
performance.
Note in this example only selects were performed.
Look to reduction in Recursive calls and Reduction in CPU time
Conclusion
Platform Version of software must be Oracle 805x now 8051 is available.
Disk configuration is extremely important. In production RAID5 is not
permitted.
Due to performance degradation. Why RAID5 will saturate your controller
throughput sooner than RAID 0+1
I did here that there was a source of tests with Oracle on Intel platform
running NT and then Linux from Orapub.
Oracle claims wild factors of 4 to 1 with Linux. Personally I don't give a
monkey's until it has been benched.
So go forth and bench it.
Peter
Plug best DBA site is without a doubt www.oriolecorp.com ask Aunt Augusta
she knows.
e.g.
Selection on the partitioned view CUSTOMER_X with index en version 7.3.4.3
select count(*) from CUSTOMER_X where billcycle = '04' and customer_id < 70001
call count cpu elapsed disk query current
rows
------- ------ -------- ---------- ---------- ---------- ----------
Parse 1 0.90 0.92 34 1692 0 0 Execute 1 0.00 0.00 0 0 0 0 Fetch 1 0.14 0.22 57 57 0 1
total 3 1.04 1.14 91 1749
0 1
Rows Execution Plan
------- ---------------------------------------------------
0 SELECT STATEMENT GOAL: CHOOSE
0 SORT (AGGREGATE)
16111 VIEW OF 'CUSTOMER_X'
16111 UNION-ALL
1 INDEX GOAL: ANALYZED (RANGE SCAN) OF 'CUST_ALL_I_01'
(NON-UNIQUE)
1 INDEX GOAL: ANALYZED (RANGE SCAN) OF 'CUST_ALL_I_02'
(NON-UNIQUE)
1 INDEX GOAL: ANALYZED (RANGE SCAN) OF 'CUST_ALL_I_03'
(NON-UNIQUE)
16112 INDEX GOAL: ANALYZED (RANGE SCAN) OF 'CUST_ALL_I_04'
(NON-UNIQUE)
1 INDEX GOAL: ANALYZED (RANGE SCAN) OF 'CUST_ALL_I_05'
(NON-UNIQUE)
1 INDEX GOAL: ANALYZED (RANGE SCAN) OF 'CUST_ALL_I_06'
(NON-UNIQUE)
1 INDEX GOAL: ANALYZED (RANGE SCAN) OF 'CUST_ALL_I_07'
(NON-UNIQUE)
1 INDEX GOAL: ANALYZED (RANGE SCAN) OF 'CUST_ALL_I_08'
(NON-UNIQUE)
1 INDEX GOAL: ANALYZED (RANGE SCAN) OF 'CUST_ALL_I_09'
(NON-UNIQUE)
Selection on the partitioned view CUSTOMER_X with index en version 8.0.5.0
select count(*)from CUSTOMER_X where billcycle = '04' and customer_id < 70001
call count cpu elapsed disk query current
rows
------- ------ -------- ---------- ---------- ---------- ----------
Parse 1 0.79 0.80 0 0 0 0 Execute 1 0.00 0.00 0 0 0 0 Fetch 2 0.11 0.30 57 57 0 1
total 4 0.90 1.10 57 57 0 1
DIGITAL Results Summary
Recursive calls reduction 92%
Temps CPU reduction 3%
PR= Physical Reads CG= Consistent Gets
DBG= Db Block Gets RC= Recursive Calls
SM= Sorts Memory
CPU=CPU Time ET =Time Elapsed
Statement PR CG DBG RC SM CPU ET Partitioned view with index v7 97 1759 1 21559 1 0.20 0.20 Partitioned view with index v8 89 1717 0 3418 29 0.59 0.59 Partitioned table (no index ) 938 938 3 1118 1 0.27 0.29 Partitioned table (Global index) 76 561 4 1611 20 0.17 0.18 Partitioned table (Local index) 68 552 4 1611 200.17 0.19
SUN Results Summary
Recursive calls reduction 91%
Temps CPU reduction 84%
Statement PR CG DBG RC SM CPU ET Partitioned view with index v7 91 1749 1 21559 1 1.04 1.14 Partitioned view with index v8 94 1759 1 3613 32 0.53 0.75 Partitioned table (no index ) 964 1352 7 1312 16 0.24 0.86 Partitioned table (Global index) 84 612 5 1805 23 0.15 0.21 Partitioned table (Local index) 75 603 5 1805 230.17 0.24
NT Results Summary
Recursive calls reduction 92%
Temps CPU reduction 87% Statement PR CG DBG RC SM CPU ET Partitioned view with index v7 86 1721 1 21225 1 1.27 1.58 Partitioned view with index v8 94 1759 1 3613 32 0.53 0.75 Partitioned table (no index ) 964 1352 7 1312 16 0.24 0.86 Partitioned table (Global index) 84 612 5 1805 23 0.15 0.21 Partitioned table (Local index) 75 603 5 1805 230.17 0.24
Bahar Leblebici wrote:
> Hi,
> I need some documentation on comparison of Oracle on NT and Unix. There
> are many articles about NT or Unix but they are not comparison. Do you
> know where can I get this information?
> Thanks in advance !.
> Bahar
Received on Wed Jan 20 1999 - 07:02:35 CST
 |
 |