Response time difference

Question

Hello:
I did 2 experiments with oracle 10g on AMD 64 bit windows 2003 server.
expt 1:
Post 10000 records, read records from 1 to 10000 in batch of 100 and note the average time .
Now with 10000 records already in the database, post another 10000 records and get records as before.
This is repeated upto 200000.
The getevents average as the records size increased by 10000 approximately increasd by 20 msecs as shown below:
records --> timetaken in msecs
0 -10000 ---> 29
10000 - 20000 --> 47
20000-30000 --> 69
30000-40000 --> 101
160000-170000 -->359
170000-180000-->384
180000-190000-->414
190000-200000-->429
Expt2:
Now with all the 200000 records already in the database table, just do the get query in interevals of 100 for every 10000 batch. Suprisingly now since the recors size is always 20000, the response time is always close to 500msecs as shown below:
records --> timetaken in msecs
0 -10000 ---> 502
10000 - 20000 --> 509
20000-30000 -->500
30000-40000 --> 500
160000-170000 -->484
170000-180000-->484
180000-190000-->476
190000-200000-->480
Wondering why is there difference when the record size is of bigger size. Any pointers related to database optimizations will be very useful.
Thanks,
Ravi

Answer 1

The sql query is :
"SELECT EVENT_ID, TYPE, VALUE, CREATE_DATE, CONTEXT_OBJ, PROCESS_INSTANCE_ID, PROCESS_TEMPLATE_ID FROM BIZEVENT EVT WHERE EVENT_ID > ? AND EVENT_ID <= ? AND ((TYPE = 'BizLogic' AND VALUE = 'W_ACTIVATED' AND PROCESS_TEMPLATE_ID = ?) OR (TYPE = 'START' AND VALUE = 'START') OR (TYPE = 'END' AND VALUE = 'END')) ORDER BY EVENT_ID");
The BizEvent table has event_id as the primarykey
The code to read a hundred block is as below:
final public static long getEvents(long start, long end, long ptid){
          long starttime = System.currentTimeMillis();
     PreparedStatement pstmt = null;
     Connection conn = null;
     try {
               //dsource is used and is already cached.
               conn = dsource.getConnection();
               pstmt = conn.prepareStatement(query);
               pstmt.setLong(1, start);
               pstmt.setLong(2, end);
               pstmt.setLong(3, ptid);
               pstmt.executeQuery();
          } catch (Exception ex) {
          ex.printStackTrace();
          } finally {
               try{
          if (pstmt != null)
               pstmt.close();
               }catch(Exception err){}
               try{
               if (conn != null)
                    conn.close();
               }catch(Exception err){}
          long finishtime = System.currentTimeMillis();
          return (finishtime - starttime);
The difference between start and end id will be 100.
Thanks for your time,