Physical reads and autotrace

Hii all
In http://asktom.oracle.com/pls/asktom/f?p=100:11:0::::P11_QUESTION_ID:880343948514 tom gives an example table
has a 320 block and the query "select * from table"
autotrace output shows physical reads=46 the statics physical reads= number of block read from disk but table has 320 block how this could be ?
Best regards..

As Oracle has a buffer cache, Oracle will place blocks in buffer cache as soon as it needs it.
When it needs those records subsequently, there will be a consistent get only.
Logical I/O is predictable. Physical I/O depends on the size of the cache.
There will never be a 1 on 1 relationship between size of a table and physical reads or you must disable the buffer cache.
Clarifications about Asktom articles are also best asked at Asktom.
Sybrand Bakker
Senior Oracle DBA
Experts: those who did read documentation.

Similar Messages

Find physical reads and logical reads ?

Hi,
how will find out physical reads and logical reads ?

Well I would suggest you read the report from statspack/awr.They as suggested by Amit,have a load profile section.That would be helpful for you in finding the details of this thing. Also which version you are? If you are in 10g than the EM is able to give you a compare period report where you can compare 2 different days's periods information and can check which particular part has changed.
In addition to this , look for the information of the metrics in the documentation.As the physical read and logical reads are statistics which are happening in the system.So from 10g onwards, oracle is keeping a track in the deflectionin the statistics. So if you see that than it willbe easy for you to manage and monitor it.
I shall try to findthe name of some views related to it and post.
Aman....

View to find Physical read and Write per hour

HI,
My oracle database is running with 11.2.0.3 and we are analysing the database. we would like to understand what the average load time per hour is for a database under normal circumstances e.g. 5 TB /Hr ?
What I want to know is, Is it possible to find the amount physical reads and writes happened per hour basis on database. Is there any view to find this ?
Thanks,
Kesav.

1e36fe40-5d07-412e-8ed2-704c647ed7a7 wrote:
HI,
My oracle database is running with 11.2.0.3 and we are analysing the database. we would like to understand what the average load time per hour is for a database under normal circumstances e.g. 5 TB /Hr ?
What I want to know is, Is it possible to find the amount physical reads and writes happened per hour basis on database. Is there any view to find this ?
Thanks,
Kesav.
number of READ operations?
number of bytes read?
SQL> desc v$IOSTAT_FILE
Name                                      Null?    Type
FILE_NO                                            NUMBER
FILETYPE_ID                                        NUMBER
FILETYPE_NAME                                      VARCHAR2(28)
SMALL_READ_MEGABYTES                               NUMBER
SMALL_WRITE_MEGABYTES                              NUMBER
LARGE_READ_MEGABYTES                               NUMBER
LARGE_WRITE_MEGABYTES                              NUMBER
SMALL_READ_REQS                                    NUMBER
SMALL_WRITE_REQS                                   NUMBER
SMALL_SYNC_READ_REQS                               NUMBER
LARGE_READ_REQS                                    NUMBER
LARGE_WRITE_REQS                                   NUMBER
SMALL_READ_SERVICETIME                             NUMBER
SMALL_WRITE_SERVICETIME                            NUMBER
SMALL_SYNC_READ_LATENCY                            NUMBER
LARGE_READ_SERVICETIME                             NUMBER
LARGE_WRITE_SERVICETIME                            NUMBER
ASYNCH_IO                                          VARCHAR2(9)
ACCESS_METHOD                                      VARCHAR2(11)
RETRIES_ON_ERROR                                   NUMBER
SQL>

Physical reads, writes and logical reads,writes

hi there,
is there any way to find the physical reads and writes, logical reads and writes
other than statspack report?

It depends which version of Oracle are you using.
If you are using 10g, you can get very comprehensive information from the dictionary views.
If you are on <=9i, then, use, v$sysstat and look for 'physical reads', 'logical reads'.
The values dervied from v$sysstat are cummulative values. i.e. values counted since the instance started.
Jaffar

No physical reads, plenty of consistant gets

Hi All,
Oracle v11.2 on Linux.
Please have a look at the query I ran and the output. The SQL_ID is of a SELECT command.
What does this situation represents, where there are zero physical reads and plenty of consistant gets ?
For consistant gets, we do read undo information (correct ??), if that undo is read from disk, will that be a "physical read"? i.e. if we read disks for consistant gets, will that be counted under physical reads or not ?
How can I describe the exact data retrival of the command here? Is it the case of "everything it needs is found in the buffer cache" ?
select a.sid, a.value , B.NAME , s.sql_id
from v$sesstat a, v$statname b, v$session s
where A.STATISTIC# = B.STATISTIC#
and b.name in ( 'redo size','physical read bytes','physical reads cache','consistent gets' )
and a.sid = s.sid
and a.sid=1018
order by a.sid;
       SID      VALUE NAME                                                             SQL_ID
      1018    7281396 consistent gets                                                  434u36htuz0s9
      1018          0 physical reads cache                                             434u36htuz0s9
      1018          0 physical read bytes                                              434u36htuz0s9
      1018       4448 redo size                                                        434u36htuz0s9
4 rows selected.Thanks in advance.

>
There are no physical reads, so weather its doing consistant-gets or not, can I say, all data required for the SELECT was in buffer cache ?
>
The data for those system views is cached in memory so Oracle does not read the disk (except at startup) to gather the information.
Some static information, like dictionary objects, is stored stored on disk in the system tablespace but this data is read when the database is mounted and stored in memory structures.
Other dynamic information, like session info, is only stored in memory structures (similar to C arrays) and Oracle can query these as if they were tables.
So no, the data was not in the buffer cache. It was already in system memory.

Logical read Vs Physical Reads

Hi,
I want to know How should we balance the logical reads and physical reads on database.
I generaly assume logical reads is always better but what if too many session accessing the same
objects it will cause latch contention.
If it is physical read then it will take more time to read from disk than reading from SGA.
If I look at statspack report I check the
Logical reads: 59,846.54 840.03
Physical reads: 1,095.91 15.38
Is there any standards for logical reads on database. If there is latch contentions can we go for the
decreasing the SGA to avoid the contentions so that there will balance between the physical reads
and logical reads.

amitbansode wrote:
I want to know How should we balance the logical reads and physical reads on database.There is no balance as it implies some kind of "+perfect ratio+" for logical I/O vs. physical I/O.
A high percentage of physical I/O can be perfectly acceptable and normal and correct for a specific database (e.g. think of a system collecting telemetry data where 90% or more of all I/O is writing new telemetry into the database and the remaining 10% is querying the data, with old data being aged out from the database using partition drops which is negligible I/O).
A high percentage of logical I/O can be indicative of a serious application design problem - where 80GB of database data is read and read again and again and again.. resulting in over a TB of logical I/O. (actually saw this in production database some years ago)
So there is no balance (e.g. not true that physical I/O = BAD and logical I/O = GOOD). No perfect cache hit ratio figure that tells you that the database is doing the right amounts of logical and physical I/O.
And I want to emphasise what Mark said - tuning requires you to identify the performance problem first, before trying to solve it.
It is very dangerous to take one metric, like the I/O cache hit ratio, and attempt to tune that. It alone is meaningless. Just like memory utilisation alone is useless and CPU utilisation alone is useless. It does not by any means point to an actual performance problem. E.g. 100% CPU utilisation can mean hardware has insufficient horses, instead of performance issues related with application design, database setting or kernel configuration.
I often repeat the following mantra here on OTN - a fundamental concept IMO for software engineering:
A solution is only as good as the problem definition.
Identify the problem first - correctly and comprehensively. And then solve it.

Physical read & logical read

Hi
What is ment by Physical-Read and Logical-Read in oracle?Can anybody explain me clearly.
Aqueel.

Physical read, reading from the disk and logical reads mean reading it from the memory.
When a session(query) required information (data blocks), it will first look into the buffer cache (SGA), if it find the required data block, it will read it from the memory, i.e. logical reads, but, if it doesn't find required data block in the memory, then it read from the disk(datafile), physical read.
Generally, people motive to reduce physical read and read as much as possible from the memory, I was also int he same wrong impression until I came across a good paper by tuning expert Cary Milsap, 'why we should focus on LIO instead of PIO'. also, search http://asktom.oracle.com 'logical reads', there are good discussion on this point.
Jaffar

What causes BUFFER GETS and PHYSICAL READS in INSERT operation to be high?

Hi All,
Am performing a huge number of INSERTs to a newly installed Oracle XE 10.2.0.1.0 on Windows. There is no SELECT statement running, but just INSERTs one after the other of 550,000 in count. When I monitor the SESSION I/O from Home > Administration > Database Monitor > Sessions, I see the following stats:
BUFFER GETS = 1,550,560
CONSISTENT GETS = 512,036
PHYSICAL READS = 3,834
BLOCK CHANGES = 1,034,232
The presence of 2 stats confuses. Though the operation is just INSERT in database for this session, why should there be BUFFER GETS of this magnitude and why should there by PHYSICAL READS. Aren't these parameters for read operations? The BLOCK CHANGES value is clear as there are huge writes and the writes change these many blocks. Can any kind soul explain me what causes there parameters to show high value?
The total columns in the display table are as follows (from the link mentioned above)
1. Status
2. SID
3. Database Users
4. Command
5. Time
6. Block Gets
7. Consistent Gets
8. Physical Reads
9. Block Changes
10. Consistent Changes
What does CONSISTENT GETS and CONSISTENT CHANGES mean in a typical INSERT operation? And does someone know which all tables are involved in getting these values?
Thank,
...

Flake wrote:
Hans, gracias.
The table just have 2 columns, both of which are varchar2 (500). No constraints, no indexes, neither foreign key references are in place. The total size of RAM in system is 1GB, and yes, there are other GUI's going on like Firefox browser, notepad and command terminals.
But, what does these other applications have to do with Oracle BUFFER GETS, PHYSICAL READS etc.? Awaiting your reply.Total RAM is 1GB. If you let XE decide how much RAM is to be allocated to buffers, on startup that needs to be shared with any/all other applications. Let's say that leaves us with, say 400M for the SGA + PGA.
PGA is used for internal stuff, such as sorting, which is also used in determing the layout of secondary facets such as indexes and uniqueness. Total PGA usage varies in size based on the number of connections and required operations.
And then there's the SGA. That needs to cover the space requirement for the data dictionary, any/all stored procedures and SQL statements being run, user security and so on. As well as the buffer blocks which represent the tablespace of the database. Since it is rare that the entire tablespace will fit into memory, stuff needs to be swapped in and out.
So - put too much space pressure on the poor operating system before starting the database, and the SGA may be squeezed. Put that space pressure on the system and you may enbd up with swapping or paging.
This is one of the reasons Oracle professionals will argue for dedicated machines to handle Oracle software.

Consistent gets and physical reads

Hi all,
I am tuning a DM SQL query, by comparing execution plans with STAR TRANSFORMATION enabled or disabled. I got the following results:
STAR TRANSFORMATION ON
74889 consistent gets
254365 physical reads
STAR TRANSFORMATION OFF
1945892 consistent gets
168028 physical reads
I thought a physical read would be counted as a logical read as well, because the data block would be read from disk (1 physical IO), placed in the buffer cache and then read from there (1 more logical IO or consistent get).
So, one physical IO does not cause a logical IO?
Thanks!
Edited by: user10634835 on 12-Jul-2011 08:40

But shouldn't consistent gets be >= physical reads (Since, as per my understanding, 1 PIO causes at least 1 LIO)? In this case it is not.
74889 consistent gets
254365 physical readsJust clarifying for my knowledge.
regards

No of physical reads of buffers from disk on PROD and Clone has huge differ

Please help !
I am having a problem interpreting a tkprof report.
I am executing the same statement below in PROD and Clone of PROD.
What I do not understand why the tkprof report shows huge difference in disk reads for same executions from PROD and Clone of PROD. I.e. first execution disk reads = 757187 , second = 6.
I expected to see Similar disk reads in the second execution also.The data it fetches is same in both the Instances.
Statment
SELECT NVL(SUM(RG_OTH.DEBIT),0),NVL(SUM(RG_OTH.CREDIT),0)
FROM
JA_IN_PLA PLA,JAI_RG_OTHERS RG_OTH,HR_ALL_ORGANIZATION_UNITS HOU,
HR_LOCATIONS LOC,JA_IN_HR_ORGANIZATION_UNITS HR_ORG WHERE 1 = 1 AND
PLA.ORGANIZATION_ID = :b1 AND TRUNC(PLA.CREATION_DATE) >= :b2 AND
TRUNC(PLA.CREATION_DATE) <= :b3 AND HOU.ORGANIZATION_ID =
PLA.ORGANIZATION_ID AND LOC.LOCATION_ID = PLA.LOCATION_ID AND
HR_ORG.ORGANIZATION_ID = PLA.ORGANIZATION_ID AND HR_ORG.LOCATION_ID =
PLA.LOCATION_ID AND TRANSACTION_ID = 19 AND INVENTORY_ITEM_ID != 0 AND
NVL(ROUNDING_ID,1) != -1 AND RG_OTH.SOURCE_REGISTER_ID = PLA.REGISTER_ID
AND RG_OTH.SOURCE_REGISTER = 'PLA' AND RG_OTH.TAX_TYPE IN (
'EXCISE_EDUCATION_CESS' )
Tkproff from PROD.
call count cpu elapsed disk query current rows
Parse 1 0.01 0.01 0 0 0 0
Execute 1 0.04 0.03 0 0 0 0
Fetch 1 973.41 1724.95 757187 789442 0 1
total 3 973.46 1725.01 757187 789442 0 1
Tkproff from CLONE of PROD.
call count cpu elapsed disk query current rows
Parse 1 0.06 0.05 1 789 0 0
Execute 1 0.00 0.00 0 0 0 0
Fetch 1 0.73 0.72 6 30594 0 1
total 3 0.79 0.78 7 31383 0 1

wins wrote:
SELECT NVL(SUM(RG_OTH.DEBIT),0),NVL(SUM(RG_OTH.CREDIT),0)
FROM
JA_IN_PLA PLA,JAI_RG_OTHERS RG_OTH,HR_ALL_ORGANIZATION_UNITS HOU,
HR_LOCATIONS LOC,JA_IN_HR_ORGANIZATION_UNITS HR_ORG WHERE 1 = 1 AND
PLA.ORGANIZATION_ID = :b1 AND TRUNC(PLA.CREATION_DATE) >= :b2 AND
TRUNC(PLA.CREATION_DATE) <= :b3 AND HOU.ORGANIZATION_ID =
PLA.ORGANIZATION_ID AND LOC.LOCATION_ID = PLA.LOCATION_ID AND
HR_ORG.ORGANIZATION_ID = PLA.ORGANIZATION_ID AND HR_ORG.LOCATION_ID =
PLA.LOCATION_ID AND TRANSACTION_ID = 19 AND INVENTORY_ITEM_ID != 0 AND
NVL(ROUNDING_ID,1) != -1 AND RG_OTH.SOURCE_REGISTER_ID = PLA.REGISTER_ID
AND RG_OTH.SOURCE_REGISTER = 'PLA' AND RG_OTH.TAX_TYPE IN (
'EXCISE_EDUCATION_CESS' )
Tkproff from PROD.
call     count       cpu    elapsed       disk      query    current        rows
Parse        1      0.01       0.01          0          0          0           0
Execute      1      0.04       0.03          0          0          0           0
Fetch        1    973.41    1724.95     757187     789442          0           1
total        3    973.46    1725.01     757187     789442          0           1
Tkproff from CLONE of PROD.
call     count       cpu    elapsed       disk      query    current        rows
Parse        1      0.06       0.05          1        789          0           0
Execute      1      0.00       0.00          0          0          0           0
Fetch        1      0.73       0.72          6      30594          0           1
total        3      0.79       0.78          7      31383          0           1
As others have pointed out, this looks like a change in execution plan. At a guess, the PROD database is doing a tablescan somewhere (physical reads is approximately equal to logical reads is a pointer).
There are many reasons why this could happen, continuing with the guesswork - we note that the statement was re-optimised (hard parsed) in the clone, but used an existing cursor in production (logical I/O = 0 in prod, 789 in clone). So the plan you got in the clone was the one that Oracle thought best for the inputs you had supplied. The plan in prod could have been produced as the response to a completely different set of input; since you have two bind variables applied to a range based predicated and one bind variable with equality it's possible that you have a histogram that makes a difference on the equality column, alternatively the range of dates used to generate the original plan was very different from the range used in the test.
Regards
Jonathan Lewis
http://jonathanlewis.wordpress.com
Author: <b><em>Oracle Core</em></b>

High user input output (I/O) and physical reads

Hi guys
Recently we have noticed that our database performance degraded significantly. as we looked around in grid control we noticed that the user IO and Physical reads are really high which cause the database to be laggy.
at first we thought it was memory problem, so we added 2gb to sga. we have set sga to be automatic and it's been like this for the past 2 years.
how do i trouble shoot this problem. i haven't found any troubleshooting guide for it yet.
please instruct me on how to solve it
database: 10.2.0.4 sparc
regards
Sina

Top 5 Timed events
db file sequential read          23,582          679          29          46.3     User I/O
CPU time                                        626                    42.7
db file scattered read     3,089     70     23     4.8     User I/O
db file parallel write     3,478     28     8     1.9     System I/O
log file parallel write     4,201     24     6     1.7     System I/O
SGA Memory Summary
SGA regions     Begin Size (Bytes)     End Size (Bytes) (if different)
Database Buffers     2,768,240,640     2,835,349,504
Fixed Size     2,050,240
Redo Buffers     14,721,024
Variable Size     3,657,439,040     3,590,330,176
Process Memory Summary
     Category     Alloc (MB)     Used (MB)     Avg Alloc (MB)     Std Dev Alloc (MB)     Max Alloc (MB)     Hist Max Alloc (MB)     Num Proc     Num Alloc
B     Other     569.22     0     2.62     3.14     22     24     217     217
     Freeable     101.63     0.00     0.77     0.45     3          132     132
     SQL     91.51     45.35     0.44     0.76     4     78     209     192
     PL/SQL     12.50     5.57     0.06     0.07     0     4     217     217
     JAVA     5.40     5.38     1.08     0.09     1     2     5     5
E     Other     603.20          3.03     3.62     22     24     199     199
     SQL     115.21     64.78     0.60     1.67     21     78     191     182
     Freeable     112.31     0.00     0.83     0.43     3          135     135
     PL/SQL     13.89     5.94     0.07     0.12     1     4     199     199
     JAVA     8.93     8.89     1.12     0.23     2     2
for some reason i can't paste a table into this post but if you want i can email you above tables

Parrallel query and physical reads

Hi I have ran a query in uing OPQ degree 12 several time, but in subsiquent runs, im still seeing physical reads, should the data not be cached after the first run..
ie, run 4 is shown below:
Statistics
         33 recursive calls
          0 db block gets
       4866 consistent gets
       3146 physical reads
          0 redo size
    7373850 bytes sent via SQL*Net to client
     293742 bytes received via SQL*Net from client
      26668 SQL*Net roundtrips to/from client
          2 sorts (memory)
          0 sorts (disk)
     400002 rows processed

http://asktom.oracle.com/pls/asktom/f?p=100:11:0::::P11_QUESTION_ID:1027523970333
parallel query may use direct io, it may not use direct io.
parallel query may use the buffer cache, it may bypass it.
And in the next followup.
Tom, in what conditions parallel query would entirely bypass buffer cache? Thanks.
Followup June 15, 2005 - 10am Central time zone:
when it does direct IO to read the blocks.
"

9582.69043 gigabytes of physical read total bytes and increasing!

In EM
Database Instance: PROD > Top Activity > I got following
physical read total bytes     62763565056     10289335500800     4183122176
cell physical IO interconnect bytes     62763565056     10289335500800     4183122176
physical read bytes     62763565056     10289335500800     4183122176
And the session is running following update procedure:
declare
FM_BBBB MT.BBBB_CODE%TYPE;
l_start NUMBER;
cursor code_upd is select /*+ parallel(FM_KWT_POP_BBBB_MISMATCH, 10) */ DDD_CID, DDD_BBBB, CCCC_BBBB from MT_MISMATCH;
begin
-- Time regular updates.
l_start := DBMS_UTILITY.get_time;
FOR rec IN code_upd LOOP
update /*+ parallel(MT, 10) nologging */ MT
set BBBB_code = rec.CCCC_BBBB
where source= 0
and cid_no = rec.DDD_CID
and BBBB_code = rec.DDD_BBBB;
commit;
END LOOP;
DBMS_OUTPUT.put_line('Bulk Updates : ' || (DBMS_UTILITY.get_time - l_start));
end;
There are 9.5 million records in MT but source=0 have only 3 million records and 376K records in MT_MISMATCH, What I don't understand why this is taking so much of time and so many bytes read? Both Tables are analyzed before running this procedure.
Can someone shed some light on this? Is there any better way of doing the same job?

Nabeel Khan wrote:
In EM
Database Instance: PROD > Top Activity > I got following
physical read total bytes     62763565056     10289335500800     4183122176
cell physical IO interconnect bytes     62763565056     10289335500800     4183122176
physical read bytes     62763565056     10289335500800     4183122176
And the session is running following update procedure:
declare
FM_BBBB MT.BBBB_CODE%TYPE;
l_start NUMBER;
cursor code_upd is select /*+ parallel(FM_KWT_POP_BBBB_MISMATCH, 10) */ DDD_CID, DDD_BBBB, CCCC_BBBB from MT_MISMATCH;
begin
-- Time regular updates.
l_start := DBMS_UTILITY.get_time;
FOR rec IN code_upd LOOP
update /*+ parallel(MT, 10) nologging */ MT
set BBBB_code = rec.CCCC_BBBB
where source= 0
and cid_no = rec.DDD_CID
and BBBB_code = rec.DDD_BBBB;
commit;
END LOOP;
DBMS_OUTPUT.put_line('Bulk Updates : ' || (DBMS_UTILITY.get_time - l_start));
end;
There are 9.5 million records in MT but source=0 have only 3 million records and 376K records in MT_MISMATCH, What I don't understand why this is taking so much of time and so many bytes read? Both Tables are analyzed before running this procedure.
Can someone shed some light on this? Is there any better way of doing the same job?Lots of badness going on here.
1) looping / procedural code where none is needed.
2) commit within the loop, one of the worst evils of all in Oracle. Please read this
http://asktom.oracle.com/pls/asktom/f?p=100:11:0::::P11_QUESTION_ID:2680799800346456179
I'd look into rewriting this as a single SQL (maybe merge) statement. Or at worst .. a bulk process utilizing collections and FORALL's.

Difference between "Physical Read Total IO Requests Per Sec" and "Physical Read IO Requests Per Sec"

What is the difference between "Physical Read Total IO Requests Per Sec" and "Physical Read IO Requests Per Sec" metrics in dba_hist_sysmetric_summary?

From docs (Reference, Statistics Descriptions) :
physical read IO requests : Number of read requests for application activity (mainly buffer cache and direct load operation) which read one or more database blocks per request. This is a subset of "physical read total IO requests" statistic.
physical read total IO requests : Number of read requests which read one or more database blocks for all instance activity including application, backup and recovery, and other utilities.

AWR's buffer cache reads and logical reads

In AWR report under "Segments by Logical Reads" section, there is a total logical reads, I assume it is in unit of block. Under "IOStat by Function summary" section, it has buffer cache reads in unit of bytes. Shouldn't the number of logical reads x 8k (if the block size is 8k) = the number of buffer cache reads?
They are not equal, not even close, does anybody know why? Thanks

Hi,
buffer gets = number of times a block was requested from buffer cache. A buffer get always request in a logical read. Depending on whether or not a copy of the block is available in the buffer cache, a logical read may or may not involve a physical read. So "buffer gets" and "logical reads" are basically synonyms and are often used interchangeably.
Oracle doesn't have a special "undo buffer". Undo blocks are stored in rollback segments in UNDO tablespace, and are managed in the same way data blocks are (they're even protected by redo). If a consistent get requires reading from UNDO tablespace, then statistics counters will show that, i.e. there will be one more consistent get in your autotrace.
For more information and some examples, see a thread at askTom:
http://asktom.oracle.com/pls/apex/f?p=100:11:0::::P11_QUESTION_ID:549546900346542976
Best regards,
Nikolay

Physical reads and autotrace

Similar Messages

Maybe you are looking for