I have very limited experience when using Oracle, and am after a rather simple query I imagine. I have a table which contains 1 million rows, Im trying to proof that compressing the data uses less space, however im not sure how to do this, based on this table creation below could someone please show me what i need to write to see the blocks used before/after?
CREATE TABLE OrderTableCompressed(OrderID, StaffID, CustomerID, TotalOrderValue)
as (select level, ceil(dbms_random.value(0, 1000)),
ceil(dbms_random.value(0,10000)),
round(dbms_random.value(0,10000),2)
from dual
connect by level <= 1000000);
ALTER TABLE OrderTableCompressed ADD CONSTRAINT OrderID_PKC PRIMARY KEY (OrderID);
--QUERY HERE THAT SHOWS BLOCKS USED/TIME TAKEN
SELECT COUNT(ORDERID) FROM OrderTableCompressed;
ALTER TABLE OrderTableCompressed COMPRESS;
--QUERY HERE THAT SHOWS BLOCKS USED/TIME TAKEN WHEN COMPRESSED
SELECT COUNT(ORDERID) FROM OrderTableCompressed;
I know how the compression works etc... its just applying the code to proove my theory. thanks for any help
--QUERY HERE THAT SHOWS BLOCKS USED
SELECT blocks, bytes/1024/1024 as MB
FROM user_segments
where segment_name = 'ORDERTABLECOMPRESSED';
Now compress the table: (Note the move. Without it you just change the attribute of the table and subsequent direct path inserts will create compressed blocks)
ALTER TABLE OrderTableCompressed MOVE COMPRESS;
Verify blocks:
--QUERY HERE THAT SHOWS BLOCKS USED TAKEN WHEN COMPRESSED
SELECT blocks, bytes/1024/1024 as MB
FROM user_segments
where segment_name = 'ORDERTABLECOMPRESSED';
Related
I have a table that stores the changes to a transaction. All the changes are captured into a table. One of the column that comes as part of the transaction can have many comma separated values. The number of occurrences cannot be predicted. Also this field is not mandatory and can have null values as well.
The total number of transactions that i have in the table is around 100M. Out of those the number of records for which the value is populated is 1M. Out of the 1M transactions the number of records for which the length of the record exceeds 4000 is ~37K.
I mentioned the length as 4000 since in my oracle table the column which would save this has been defined as varchar2(4000).
I check at places and found that if I have to save something of unknown length then i should define the table column datatype as clob. But clob is expensive for me since only a very small amount of data has length > 4000. If I snowflake my star schema and create another table to store the values then even though, I have transactions for which the length is much smaller than 4000 would be saved as part of the clob column. This would be expensive both in terms of storage and performance.
Can someone suggest me an approach to solve this problem.
Thanks
S
You could create a master - detail table to store the comma separated values, then you could have rows rather than save all comma separated values in a single column. This could be managed with a foregn key using a pseudo key between master and detail table.
Here's one option.
Create two columns, e.g.
create table storage
(id number primary key,
long_text_1 varchar2(4000),
long_text_2 varchar2(4000)
);
Store values like
insert into storage (id, long_text_1, long_text_2)
values (seq.nextval,
substr(input_value, 1, 4000),
substr(input_value, 4001, 4000)
);
When retrieving them from the table, concatenate them:
select id,
long_text_1 || long_text_2 as long_text
from storage
where ...
You might benefit from using inline SecurFile CLOBs. With inline CLOBs, up to about 4000 bytes of data can be stored in rows like a regular VARCHAR2 and only the larger values will be stored in a separate CLOB segment. With SecureFiles, Oracle can significantly improve CLOB performance. (For example, import and export of SecureFiles is much faster than the old-fashioned BasicFile LOB format.)
Depending on your version, parameters, and table DDL, your database may already store CLOBs as inline SecureFiles. Ensure that your COMPATIBLE setting is 11.2 or higher, and that DB_SECUREFILE is one of "permitted", "always", or "preferred":
select name, value from v$parameter where name in ('compatible', 'db_securefile') order by 1;
Use a query like this to ensure that your tables were setup correctly, and nobody overrode the system settings:
select dbms_metadata.get_ddl('TABLE', 'YOUR_TABLE_NAME') from dual;
You should see something like this in the results:
... LOB ("CLOB_NAME") STORE AS SECUREFILE (... ENABLE STORAGE IN ROW ...) ...
One of the main problems with CLOBs is that they are stored in a separate segment, and a LOB index must be traversed to map each row in the table to a value in another segment. The below demo creates two tables to show that LOB segments do not need to be used when the the data is small and stored inline.
--drop table clob_test_inline;
--drop table clob_test_not_in;
create table clob_test_inline(a number, b clob) lob(b) store as securefile (enable storage in row);
create table clob_test_not_in(a number, b clob) lob(b) store as (disable storage in row);
insert into clob_test_inline select level, lpad('A', 900, 'A') from dual connect by level <= 10000;
insert into clob_test_not_in select level, lpad('A', 900, 'A') from dual connect by level <= 10000;
commit;
The inline table segment is large, because it holds all the data. The out of line table segment is small, because all of its data is held elsewhere.
select segment_name, bytes/1024/1024 mb_inline
from dba_segments
where segment_name like 'CLOB_TEST%'
order by 1;
SEGMENT_NAME MB_INLINE
---------------- ---------
CLOB_TEST_INLINE 27
CLOB_TEST_NOT_IN 0.625
Looking at the LOB segments, the sizes are reversed. The inline table doesn't store anything in the LOB segment.
select table_name, bytes/1024/1024 mb_out_of_line
from dba_segments
join dba_lobs
on dba_segments.owner = dba_lobs.owner
and dba_segments.segment_name = dba_lobs.segment_name
where dba_lobs.table_name like 'CLOB_TEST%'
order by 1;
TABLE_NAME MB_OUT_OF_LINE
------------ --------------
CLOB_TEST_INLINE 0.125
CLOB_TEST_NOT_IN 88.1875
Despite the above, I can't promise that CLOBs will still work for you. All I can say is that it's worth testing the data using CLOBs. You'll still need to look out for a few things. CLOBs store text slightly differently (UCS2 instead of UTF8), which may take up more space depending on your character sets. So check the segment sizes. But also beware that segment sizes can lie when they are small - there's a lot of auto-allocated overhead for sample data, so you'll want to use realistic sizes when testing.
Finally, as Raul pointed out, storing non-atomic values in a field is usually a terrible mistake. That said, there are rare times when data warehouses need to break the rules for performance, and data needs to be stored as compactly as possible. Before you store the data this way, ensure that you will never need to join based on those values, or query for individual values. If you think dealing with 100M rows is tough, just wait until you try to split 100M values and then join them to another table.
Objective
Verify if it is true that insert records without PK/index plus create thme later is faster than insert with PK/Index.
Note
The point here is not about indexing takes more time (it is obvious), but the total cost (Insert without index + create index) is higher than (Insert with index). Because I was taught to insert without index and create index later as it should be faster.
Environment
Windows 7 64 bit on DELL Latitude core i7 2.8GHz 8G memory & SSD HDD
Oracle 11G R2 64 bit
Background
I was taught that insert records without PK/Index and create them after insert would be faster than insert with PK/Index.
However 1 million record inserts with PK/Index was actually faster than creating PK/Index later, approx 4.5 seconds vs 6 seconds, with the experiments below. By increasing the records to 3 million (999000 -> 2999000), the result was the same.
Conditions
The table DDL is below. One bigfile table space for both data and
index.
(Tested a separate index tablespace with the same result & inferior overall perforemace)
Flush the buffer/spool before each run.
Run the experiment 3 times each and made sure the results
were similar.
SQL to flush:
ALTER SYSTEM CHECKPOINT;
ALTER SYSTEM FLUSH SHARED_POOL;
ALTER SYSTEM FLUSH BUFFER_CACHE;
Question
Would it be actually true that "insert witout PK/Index + PK/Index creation later" is faster than "insert with PK/Index"?
Did I make mistakes or missed some conditions in the experiment?
Insert records with PK/Index
TRUNCATE TABLE TBL2;
ALTER TABLE TBL2 DROP CONSTRAINT PK_TBL2_COL1 CASCADE;
ALTER TABLE TBL2 ADD CONSTRAINT PK_TBL2_COL1 PRIMARY KEY(COL1) ;
SET timing ON
INSERT INTO TBL2
SELECT i+j, rpad(TO_CHAR(i+j),100,'A')
FROM (
WITH DATA2(j) AS (
SELECT 0 j FROM DUAL
UNION ALL
SELECT j+1000 FROM DATA2 WHERE j < 999000
)
SELECT j FROM DATA2
),
(
WITH DATA1(i) AS (
SELECT 1 i FROM DUAL
UNION ALL
SELECT i+1 FROM DATA1 WHERE i < 1000
)
SELECT i FROM DATA1
);
commit;
1,000,000 rows inserted.
Elapsed: 00:00:04.328 <----- Insert records with PK/Index
Insert records without PK/Index and create them after
TRUNCATE TABLE TBL2;
ALTER TABLE &TBL_NAME DROP CONSTRAINT PK_TBL2_COL1 CASCADE;
SET TIMING ON
INSERT INTO TBL2
SELECT i+j, rpad(TO_CHAR(i+j),100,'A')
FROM (
WITH DATA2(j) AS (
SELECT 0 j FROM DUAL
UNION ALL
SELECT j+1000 FROM DATA2 WHERE j < 999000
)
SELECT j FROM DATA2
),
(
WITH DATA1(i) AS (
SELECT 1 i FROM DUAL
UNION ALL
SELECT i+1 FROM DATA1 WHERE i < 1000
)
SELECT i FROM DATA1
);
commit;
ALTER TABLE TBL2 ADD CONSTRAINT PK_TBL2_COL1 PRIMARY KEY(COL1) ;
1,000,000 rows inserted.
Elapsed: 00:00:03.454 <---- Insert without PK/Index
table TBL2 altered.
Elapsed: 00:00:02.544 <---- Create PK/Index
Table DDL
CREATE TABLE TBL2 (
"COL1" NUMBER,
"COL2" VARCHAR2(100 BYTE),
CONSTRAINT "PK_TBL2_COL1" PRIMARY KEY ("COL1")
) TABLESPACE "TBS_BIG" ;
The current test case is probably good enough for you to overrule the "best practices". There are too many variables involved to make a blanket statement that "it's always best to leave the indexes enabled". But you're probably close enough to say it's true for your environment.
Below are some considerations for the test case. I've made this a community wiki in the hopes that others will add to the list.
Direct-path inserts. Direct-path writes use different mechanisms and may work completely differently. Direct-path inserts can often be significantly faster than regular inserts, although they have some complicated restrictions (for example, triggers must be disabled) and disadvantages (the data is not immediately backed-up). One particular way it affects this scenario is that NOLOGGING for indexes only applies during index creation. So even if a direct-path insert is used, an enabled index will always generate REDO and UNDO.
Parallelism. Large insert statements often benefit from parallel DML. Usually it's not worth worrying about the performance of bulk loads until it takes more than several seconds, which is when parallelism starts to be useful.
Bitmap indexes are not meant for large DML. Inserts or updates to a table with a bitmap index can lock the whole table and lead to disastrous performance. It might be helpful to limit the test case to b-tree indexes.
Add alter system switch logfile;? Log file switches can sometimes cause performance issues. The tests would be somewhat more consistent if they all started with empty logfiles.
Move data generation logic into a separate step. Hierarchical queries are useful for generating data but they can have their own performance issues. It might be better to create in intermediate table to hold the results, and then only test inserting the intermediate table into the final table.
It's true that it is faster to modify a table if you do not also have to modify one or more indexes and possibly perform constraint checking as well, but it is also largely irrelevant if you then have to add those indexes. You have to consider the complete change to the system that you wish to effect, not just a single part of it.
Obviously if you are adding a single row into a table that already contains millions of rows then it would be foolish to drop and rebuild indexes.
However, even if you have a completely empty table into which you are going to add several million rows it can still be slower to defer the indexing until afterwards.
The reason for this is that such an insert is best performed with the direct path mechanism, and when you use direct path inserts into a table with indexes on it, temporary segments are built that contain the data required to build the indexes (data plus rowids). If those temporary segments are much smaller than the table you have just loaded then they will also be faster to scan and to build the indexes from.
the alternative, if you have five index on the table, is to incur five full table scans after you have loaded it in order to build the indexes.
Obviously there are huge grey areas involved here, but well done for:
Questioning authority and general rules of thumb, and
Running actual tests to determine the facts in your own case.
Edit:
Further considerations -- you run a backup while the indexes are dropped. Now, following an emergency restore, you have to have a script that verifies that all indexes are in place, when you have the business breathing down your neck to get the system back up.
Also, if you absolutely were determined to not maintain indexes during a bulk load, do not drop the indexes -- disable them instead. This preserves the metadata for the indexes existence and definition, and allows a more simple rebuild process. Just be careful that you do not accidentally re-enable indexes by truncating the table, as this will render disabled indexes enabled again.
Oracle has to do more work while inserting data into table having an index. In general, inserting without index is faster than inserting with index.
Think in this way,
Inserting rows in a regular heap-organized table with no particular row order is simple. Find a table block with enough free space, put the rows randomly.
But, when there are indexes on the table, there is much more work to do. Adding new entry for the index is not that simple. It has to traverse the index blocks to find the specific leaf node as the new entry cannot be made into any block. Once the correct leaf node is found, it checks for enough free space and then makes the new entry. If there is not enough space, then it has to split the node and distribute the new entry into old and new node. So, all this work is an overhead and consumes more time overall.
Let's see a small example,
Database version :
SQL> SELECT banner FROM v$version where ROWNUM =1;
BANNER
--------------------------------------------------------------------------------
Oracle Database 12c Enterprise Edition Release 12.1.0.1.0 - 64bit Production
OS : Windows 7, 8GB RAM
With Index
SQL> CREATE TABLE t(A NUMBER, CONSTRAINT PK_a PRIMARY KEY (A));
Table created.
SQL> SET timing ON
SQL> INSERT INTO t SELECT LEVEL FROM dual CONNECT BY LEVEL <=1000000;
1000000 rows created.
Elapsed: 00:00:02.26
So, it took 00:00:02.26. Index details:
SQL> column index_name format a10
SQL> column table_name format a10
SQL> column uniqueness format a10
SQL> SELECT index_name, table_name, uniqueness FROM user_indexes WHERE table_name = 'T';
INDEX_NAME TABLE_NAME UNIQUENESS
---------- ---------- ----------
PK_A T UNIQUE
Without Index
SQL> DROP TABLE t PURGE;
Table dropped.
SQL> CREATE TABLE t(A NUMBER);
Table created.
SQL> SET timing ON
SQL> INSERT INTO t SELECT LEVEL FROM dual CONNECT BY LEVEL <=1000000;
1000000 rows created.
Elapsed: 00:00:00.60
So, it took only 00:00:00.60 which is faster compared to 00:00:02.26.
Is it possible to have a table that reports a size but does not have any rows in it? When I run the following query one of the tables reports a size but does not contain any rows. How is this possible?
select table_name,
b.tablespace_name,
sum( bytes)/1024/1024 "SIZE IN MB"
from USER_segments a,
user_tables b
where table_name=segment_name
group by segment_name,
b.tablespace_name,
table_name;
Table segements grow when data is inserted into them. Since 11g a new created table can be created without a segment. When data is inserted into such a newly created table, the segment is created.
The space that is occupied by the segment is not automatically returned to the free space in the datafile when rows are deleted. So, your table is created empty with direct segment creation, or it had rows and they are deleted.
I currently have a Hive table that has 1.5 billion rows. I would like to create a smaller table (using the same table schema) with about 1 million rows from the original table. Ideally, the new rows would be randomly sampled from the original table, but getting the top 1M or bottom 1M of the original table would be ok, too. How would I do this?
As climbage suggested earlier, you could probably best use Hive's built-in sampling methods.
INSERT OVERWRITE TABLE my_table_sample
SELECT * FROM my_table
TABLESAMPLE (1m ROWS) t;
This syntax was introduced in Hive 0.11. If you are running an older version of Hive, you'll be confined to using the PERCENT syntax like so.
INSERT OVERWRITE TABLE my_table_sample
SELECT * FROM my_table
TABLESAMPLE (1 PERCENT) t;
You can change the percentage to match you specific sample size requirements.
You can define a new table with the same schema as your original table.
Then use INSERT OVERWRITE TABLE <tablename> <select statement>
The SELECT statement will need to query your original table, use LIMIT to only get 1M results.
This query will pull out top 1M rows and overwrite them in a new table.
CREATE TABLE new_table_name AS
SELECT col1, col2, col3, ....
FROM original_table
WHERE (if you want to put any condition) limit 100000;
I am facing problem in loading data. I have to copy 800,000 rows from one table to another in Oracle database.
I tried for 10,000 rows first but the time it took is not satisfactory. I tried using the "BULK COLLECT" and "INSERT INTO SELECT" clause but for both the cases response time is around 35 minutes. This is not the desired response I'm looking for.
Does anyone have any suggestions?
Anirban,
Using an "INSERT INTO SELECT" is the fastest way to populate your table. You may want to extend it with one or two of these hints:
APPEND: to use direct path loading, circumventing the buffer cache
PARALLEL: to use parallel processing if your system has multiple cpu's and this is a one-time operation or an operation that takes place at a time when it doesn't matter that one "selfish" process consumes more resources.
Just using the append hint on my laptop copies 800,000 very small rows below 5 seconds:
SQL> create table one_table (id,name)
2 as
3 select level, 'name' || to_char(level)
4 from dual
5 connect by level <= 800000
6 /
Tabel is aangemaakt.
SQL> create table another_table as select * from one_table where 1=0
2 /
Tabel is aangemaakt.
SQL> select count(*) from another_table
2 /
COUNT(*)
----------
0
1 rij is geselecteerd.
SQL> set timing on
SQL> insert /*+ append */ into another_table select * from one_table
2 /
800000 rijen zijn aangemaakt.
Verstreken: 00:00:04.76
You mention that this operation takes 35 minutes in your case. Can you post some more details, so we can see what exactly is taking 35 minutes?
Regards,
Rob.
I would agree with Rob. Insert into () select is the fastest way to do this.
What exactly do you need to do? If you're trying to do a table rename by copying to a new table and then deleting the old, you might be better off doing a table rename:
alter table
table
rename to
someothertable;
INSERT INTO SELECT is the fastest way to do it.
If possible/necessary, disable all indexes on the target table first.
If you have no existing data in the target table, you can also try CREATE AS SELECT.
As with the above, I would recommend the Insert INTO ... AS select .... or CREATE TABLE ... AS SELECT ... as the fastest way to copy a large volume of data between two tables.
You want to look up the direct-load insert in your oracle documentation. This adds two items to your statements: parallel and nologging. Repeat the tests but do the following:
CREATE TABLE Table2 AS SELECT * FROM Table1 where 1=2;
ALTER TABLE Table2 NOLOGGING;
ALTER TABLE TABLE2 PARALLEL (10);
ALTER TABLE TABLE1 PARALLEL (10);
ALTER SESSION ENABLE PARALLEL DML;
INSERT INTO TABLE2 SELECT * FROM Table 1;
COMMIT;
ALTER TABLE 2 LOGGING:
This turns off the rollback logging for inserts into the table. If the system crashes, there's not recovery and you can't do a rollback on the transaction. The PARALLEL uses N worker thread to copy the data in blocks. You'll have to experiment with the number of parallel worker threads to get best results on your system.
Is the table you are copying to the same structure as the other table? Does it have data or are you creating a new one? Can you use exp/imp? Exp can be give a query to limit what it exports and then imported into the db. What is the total size of the table you are copying from? If you are copying most of the data from one table to a second, can you instead copy the full table using exp/imp and then remove the unwanted rows which would be less than copying.
try to drop all indexes/constraints on your destination table and then re-create them after data load.
use /*+NOLOGGING*/ hint in case you use NOARCHIVELOG mode, or consider to do the backup right after the operation.