SQL*Net version 1 and 2
From: Tony Cook <TCook_at_vmcsnat4.telecom.com.au>
Date: 1995/06/20
Message-ID: <3s5j0h$do_at_newsserver.trl.OZ.AU>#1/1
Tony Cook
Telecom Australia
Email: tcook_at_vmcsnat4.telecom.com.au Received on Tue Jun 20 1995 - 00:00:00 CEST
Date: 1995/06/20
Message-ID: <3s5j0h$do_at_newsserver.trl.OZ.AU>#1/1
A quick comparison between version 1 & 2 of SQL*Net running on a 486SX/33 to an HP E45 shows the following approximate times to perform a CONNECT.
- Version 1 5 seconds
- Version 2 with Name lookup 19 seconds
- Version 2 full connect string 18 seconds
(The results are averages of 3-4 iterations of the test)
For each test the LOCAL variable in the ORACLE.INI was set to a string of the form:
1. T:hostname:SID 2. TNS:SID.domain.com.au 3. (DESCRIPTION=(ADDRESS=(PROTOCOL=TCP)(HOST=hostname.domain.com.au) (PORT=1521))(CONNECT_DATA=(SRVR=DEDICATED)(SID=SID)))
Notes:
- The host is running an Oracle Names server.
- No difference in timing was noted if the test 3 connect string used SRVR=SHARED to connect via an MTS connection.
- I assume that the name lookup is not performed in test 3, saving about 1 second.
These results are particularly worrying as version 1 is to be desupported as of the end of this year.
Does anyone have any idea as to why the overhead on version 2 is so large? Can I expect that once connected version 2 is as fast as version 1? Why should I even consider version 2?
Any thoughts you may have would be appreciated.
Thanks,
Tony Cook
Tony Cook
Telecom Australia
Email: tcook_at_vmcsnat4.telecom.com.au Received on Tue Jun 20 1995 - 00:00:00 CEST