Metrics analysis

Question

Hi All,
System - Oracle 10.2.0.4 on SOLARIS 10. Application is JDE..(Lot of adhoc queries, lot of batch/ETL jobs at Night, normal OLTP)
Using RAID 1+0
In my AWR report for some time I notice the following;
1> High db file parallel read for past few months
2> High log file switch completion
3> Very low Buffer Hit Ratio from 38-55 %
Can anyone please explain on this; What needs to be done ?
AWR Report and Findings as below; Have posted only the high values
Instance Efficiency Percentages (Target 100%)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
            Buffer Nowait %:  100.00       Redo NoWait %:  100.00
            Buffer  Hit   %:   55.66    In-memory Sort %:  100.00
            Library Hit   %:   98.34        Soft Parse %:   99.25
         Execute to Parse %:   57.56         Latch Hit %:   99.97
Parse CPU to Parse Elapsd %:   90.02     % Non-Parse CPU:   98.34
Top 5 Timed Events                                         Avg %Total
~~~~~~~~~~~~~~~~~~                                        wait   Call
Event                                 Waits    Time (s)   (ms)   Time Wait Class
CPU time                                          4,472          67.5
db file parallel read                 8,003         166     21    2.5   User I/O
                                                                   Avg
                                             %Time  Total Wait    wait     Waits
Event                                 Waits  -outs    Time (s)    (ms)      /txn
db file parallel read                 8,003     .0         166      21       0.0
lSQL*Net more data to client         834,185     .0          52       0       4.0
SQL*Net more data from clien        186,379     .0          16       0       0.9
log file sequential read                423     .0           4       8       0.0
cursor: pin S wait on X                  50  100.0           0      10       0.0
SGA: allocation forcing comp             39   79.5           0      10       0.0
log file switch completion                9     .0           0      35       0.0
latch free                               11     .0           0       8       0.0
latch: shared pool                       11     .0           0       7       0.0
SQL*Net message from client       2,691,087     .0   1,471,156     547      12.8
Streams AQ: qmn slave idle w            385     .0       7,065   18351       0.0
virtual circuit status                  121  100.0       3,542   29273       0.0
Streams AQ: qmn coordinator             264   50.4       3,532   13380       0.0
Streams AQ: waiting for mess            721  100.0       3,520    4882       0.0
wait for unread message on b          3,601  100.0       3,516     976       0.0
jobq slave wait                       1,192   98.9       3,473    2914       0.0
Streams AQ: waiting for time              4  100.0       1,058  264428       0.0
class slave wait                          2     .0           0       0       0.0
Background Wait Events             DB/Inst: HQJDDB/hqjddb  Snaps: 66170-66171
-> ordered by wait time desc, waits desc (idle events last)
                                                                   Avg
                                             %Time  Total Wait    wait     Waits
Event                                 Waits  -outs    Time (s)    (ms)      /txn
os thread startup                        48     .0           4      83       0.0
log file sequential read                423     .0           4       8       0.0
control file parallel write           1,808     .0           3       2       0.0
latch: shared pool                        1     .0           0      28       0.0
log file single write                     6     .0           0       4       0.0
lStreams AQ: qmn slave idle w            385     .0       7,065   18351       0.0
Streams AQ: qmn coordinator             264   50.4       3,532   13380       0.0
smon timer                              317     .0       3,507   11064       0.0
pmon timer                            1,689  100.0       3,499    2071       0.0
Streams AQ: waiting for time              4  100.0       1,058  264428       0.0
init.ora Parameters              
                                                                End value
Parameter Name                Begin value                       (if different)
_addm_auto_enable             TRUE
_optimizer_mjc_enabled        FALSE
_spin_count                   6000
db_32k_cache_size             4294967296
db_block_size                 8192
db_keep_cache_size            1073741824
db_recovery_file_dest_size    2147483648
db_writer_processes           4
filesystemio_options          ASYNCH
hash_area_size                20971520
job_queue_processes           10
large_pool_size               83886080
log_buffer                    14242816
max_dump_file_size            UNLIMITED
open_cursors                  300
parallel_adaptive_multi_user  FALSE
parallel_execution_message_si 8192
parallel_max_servers          640
parallel_threads_per_cpu      2
pga_aggregate_target          3221225472
processes                     800
recyclebin                    ON
remote_login_passwordfile     EXCLUSIVE
session_cached_cursors        300
sga_max_size                  23622320128
sga_target                    23622320128
undo_management               AUTO
My log switches over past 2 weeks;;
Date      Day 00:00 1am 2am 3am 4am 5am 6am 7am 8am 9am 10am 11am 12:00 1pm 2pm 3pm 4pm 5pm 6pm 7pm 8pm 9pm 10pm 11pm
02-MAY-12 Wed     1   2  11   6   4   1   1   1   1   2    3    7     1
01-MAY-12 Tue     2   4  13   6   5   1   1           1               3               1   1   2   1       2         1
30-APR-12 Mon     1   2  10   8   3   1   1       2   3    2    6     2   1   2   2   2   4   2   2   1   2    1
29-APR-12 Sun     2   2  12   6   5   1   1       2   1               3   2   2   2       1   2   1   1   1    1
28-APR-12 Sat     1   3  15   4   3   8   5   2   1   2    2    4     3   1   2   1   2   1   3   2       4    1
27-APR-12 Fri     1   4  14   4   3   1   1   2   1   3    6    8     4   1   3   2   2   2   2   1       2    2
26-APR-12 Thu     2   3  11   5   5       1   1      16    6    4     2   1   1   5   5   5   2   1   1   2    1
25-APR-12 Wed     2   2  14   4   5       1   1       3    2    2     2   1   1   1   1   2   2   1   1   2    1
24-APR-12 Tue     2   5  12   5   4       1   1       2    2    2     3   1   4   2   2   2   2   1   2   2    1
23-APR-12 Mon     1   2  10   4   2   1   1       1   2    2    4     6   3   2   2   2   2   2   2   3   2    1
22-APR-12 Sun     2   2  12   4   2   1   1   1  39             1     1   2   2   1       2   1   1   1   1    1
21-APR-12 Sat     2   2  13   4   3   7   5   2   1   1    3    4     3   1       2   1   2   2   2   1   2    1
20-APR-12 Fri     1   3  11   3   3   2   1   1       2    3    1     2   1   4   2   3   3   2   1   2   2    1
19-APR-12 Thu     2   4  14   3   4       1   1       1    3    1     2       2   2   1   8   3   1       2    1
18-APR-12 Wed     2   2  11   3   4       1   1   1   2    3    6     8   1   2   2   3   2   2   1   1   2    1
My redo log file sizes;
SQL> select group#, members, bytes/1024/1024 "Size (MB)" from v$log;
    GROUP#    MEMBERS  Size (MB)
         1          1        150
         2          1        150
         3          1        150
         4          1        150

Answer 1

Iordan Iotzov wrote:
1> High db file parallel read for past few monthsWhat are the criteria for “high”? This event could show up during recovery and buffer pre-fetching. At 2.5% of your total call time, why do you think this event presents a systemic issue?
In general the response time has decreased. Is there any way to reduce the wait..
Did not find much info on it from Net..
2> High log file switch completion The report shows 9 waits each lasting around 35 ms., so I do not see this as a serious performance issue at this time. Many would find 38 log switches per hour too high. You could reduce that number by increasing the size of the redo log files. From the report I see, this would likely not help your performance now, but it could be a good preventive measure.
Thinking on the same line. But how to set the optimal size of redo logs..?
SQL> select optimal_logfile_size from v$instance_recovery;
OPTIMAL_LOGFILE_SIZE
-- NO DATA.
3> Very low Buffer Hit Ratio from 38-55 %
Buffer hit ratio is pretty much meaningless. This Jonathan Lewis’ blog entry sums it up nicely - http://jonathanlewis.wordpress.com/2007/09/05/hit-ratios-2/
Concur with you. But was looking into potential problems, since the BCHR decreased from a earlier value of 65-75%