From daemon Fri Feb 2 10:22:50 1996 Received: from ccvm.sunysb.edu by alice.jcc.com; (5.65v3.0/1.1.8.2/01Aug94-0142PM) id AA28258; Fri, 2 Feb 1996 10:22:46 -0500 Message-Id: <9602021522.AA28258@alice.jcc.com> Received: from CCVM.SUNYSB.EDU by ccvm.sunysb.edu (IBM VM SMTP V2R3) with BSMTP id 3398; Fri, 02 Feb 96 10:04:35 EST Received: from CCVM.SUNYSB.EDU (NJE origin LISTSERV@SBCCVM) by CCVM.SUNYSB.EDU (LMail V1.2a/1.8a) with BSMTP id 1211; Fri, 2 Feb 1996 10:04:31 -0500 Date: Fri, 2 Feb 1996 09:45:57 -0500 Reply-To: "ORACLE database mailing list." Sender: "ORACLE database mailing list." From: Chuck Isaacson Subject: Re: How to speed up parsing? X-To: ORACLE-L@CCVM.SUNYSB.EDU To: Multiple recipients of list ORACLE-L In-Reply-To: <199602012138.QAA16877@mtu.edu> from "TSAWMILL.US.ORACLE.COM" at Feb 1, 96 04:23:01 pm Yup, wild it is, BUT, under level 6 I had a situation where the sequence of creating the indexes, there were 2, determined whether an index was used or not. When not used a full table scan was performed slowing performance to a crawl. When the creation of the indexes was reversed the full table scan was eliminated and performance was acceptable. Was told by our vendor that the situation did NOT occur in level 7. So maybe your idea is not wild. Chuck > > --Boundary-8172845-0-0 > > Also, here's another wild idea: were the indexes on the two instances created > in the same chronological order? Check the timestamps in user_objects. > > > Tim Sawmiller > tsawmiller@us.oracle.com > "The opinions expressed here are my own and not necessarily those of the > Oracle Corporation". > > > --Boundary-8172845-0-0 > Content-Type: message/rfc822 > > Date: 01 Feb 96 14:11:22 > From:"Jeffrey Weiss " > To: Multiple,recipients,of,list,ORACLE-L,ORACLE-L@CCVM.SUNYSB.EDU > Subject: Re: How to speed up parsing? > Reply-to: ORACLE-L@CCVM.SUNYSB.EDU > X-Orcl-Application: Sender: "ORACLE database mailing list." > > X-To: ORACLE-L@ccvm.sunysb.edu > > > OK. Here's more info. The two instances are both on the same machine. The > views are defined exactly the same. The indexes are the same on both > instances. I have verified this by doing set operations using database > links between the instances. The use of the instances is a little different, > one being used by developers the other by an integrated testing team. > > Each instance has its datafiles spread across several disks in a Sparc > Storage Array. Two instances do not share the same devices so there should > not be contention there. > > Both of these instances are copies (extract/import) of our production > database. It's just that the copies were made at different times. One > was made when we first went production. The other was refreshed after > another module went into production. But setups (for PA) were done > manually in the older instance. > > Jeff Weiss > > > > Subj: Re: How to speed up parsing? > > --Boundary-8166969-0-0 > > First of all, "parsing" is where Oracle deciphers the SQL provided to figure > out what it has to do. Parsing is NOT the execution of the SQL. > Secondly, we need more info. Are there indexes on both systems? Are the > views defined exactly the same way? What are the load differences of the > machines? What is the disk spread of the two machines? etc. etc. > > > Tim Sawmiller > tsawmiller@us.oracle.com > "The opinions expressed here are my own and not necessarily those of the > Oracle Corporation". > > > --Boundary-8172845-0-0-- >