Why does a SELECT query needs REDO?

From: Jining Han <jining.han_at_gmail.com>
Date: 9 Feb 2006 08:05:19 -0800
Message-ID: <1139501119.613499.283110@o13g2000cwo.googlegroups.com>

In a 10.1.0.4 2-node RAC on HP-UX Itanuim, I have a database running which doesn't seem to have any DML activities. Now I run a long running SQL query (with SUM, GROUP BY), and the AUTOTRACE stats looks like this:

Statistics

---

        482  recursive calls
          3  db block gets

    3932472 consistent gets
    3831470 physical reads
    7749080 redo size

       2486  bytes sent via SQL*Net to client
        686  bytes received via SQL*Net from client
          1  SQL*Net roundtrips to/from client
          9  sorts (memory)
          0  sorts (disk)
          0  rows processed

I am wondering why there's 7749080 redo size (physical read, I assume) involved. Received on Thu Feb 09 2006 - 10:05:19 CST