Presentation is loading. Please wait.

Presentation is loading. Please wait.

Q UESTIONS ON B ACKUP AND R ECOVERY 1. What is the value of cold backup? 2. What's the difference between startup and startup mount? When we do "alter.

Similar presentations


Presentation on theme: "Q UESTIONS ON B ACKUP AND R ECOVERY 1. What is the value of cold backup? 2. What's the difference between startup and startup mount? When we do "alter."— Presentation transcript:

1 Q UESTIONS ON B ACKUP AND R ECOVERY 1. What is the value of cold backup? 2. What's the difference between startup and startup mount? When we do "alter database open, what does it mean? 3. What's the archive log function? 4. When I did 'alter database open', I still get an error. But once I issued 'recover database' and then 'alter database open, everything works just fine. What happens between these steps? 1 ITEC 450 Fall 2012

2 M ODULE 4 D ATABASE T UNING Section 5 Optimization of Physical Design 2 ITEC 450 Fall 2012

3 T ECHNIQUES FOR O PTIMIZING D ATABASE Partitioning – breaking a single large table into sections (partitions) stored in multiple files Raw versus file systems – OS-controlled file or not Clustering – enforcing the physical sequence of data on disk Free space and compression – storage management File placement and allocation Page size (block size in Oracle) – using the proper page size for efficient data storage and I/O Fall 2012 3 ITEC 450

4 P ARTITIONING Partitioning supports very large tables and indexes by decomposing them into smaller and more manageable pieces called partitions. Fall 2012 4 ITEC 450 SQL queries and DML statements do not need to be modified in order to access partitioned tables After partitions are defined, DDL statement can access and manipulate individual partitions if they choose to, rather than entire tables or indexes

5 R AW D EVICE VS. F ILE S YSTEM Raw device is a UNIX feature to bypass operating system cache mechanism. File system is a common way to manage storage. Raw device might provide a slight performance benefit, but the database storage management is a nightmare. Unless it is required by DBMS features, you should avoid to implementing Raw Device. Fall 2012 5 ITEC 450

6 C LUSTERING Clustering is used to store one or more tables physically together with the share common columns and are often used together. Because related rows are physically stored together, disk access time improves. Clustering usually is enforced by the DBMS with a cluster index. Like indexes, clusters do not affect application design. Data stored in a clustered table is accessed by SQL in the same way as data stored in a non-clustered table. Fall 2012 6 ITEC 450

7 C LUSTERING E XAMPLE Fall 2012 7 ITEC 450

8 F REE S PACE AND C OMPRESSION Free space – pre-allocate storage of tablespace, tables and indexes to store newly added data In DDL statements, a parameter is PCTFREE Benefits and disadvantages are listed in the textbook Compression – shrink the size of a database Tables or indexes can be algorithmically compressed. The DMBS will handle the reading and writing automatically. Fall 2012 8 ITEC 450

9 F ILE P LACEMENT AND A LLOCATION A database is very I/O intensive, the location of the files can be critical to database performance and recovery. Separate the indexes from the data – traditional approach Analyze the access patterns of your applications and separate the files for tables/indexes that are frequently accessed together Place transaction logs on a separate disk device from the actual data, and spread out log members (redo01.log, redo02.log, redo03.log) to different physical disk devices Place control files (control01.ctl, control02.ctl, control03.ctl) like transaction logs to different physical disk devices. Its common to have control01.ctl shares the same device with redo01.log, etc. Place archive logs on a complete separate disk device from any database files Fall 2012 9 ITEC 450

10 B LOCK S IZE The block or page size is the smallest unit of I/O, and used to store data records. The size selection is depending on record lengths – small block size is good for small record length; large block size is good for long record length Typical page size are 2K, 4K, 8K, 16K, and 32K. Common used sizes are 2K and 4K Data Warehouse or Data Mart applications use 8K, 16K or 32K. Oracle allows to specify block size on tablespace level – tablespace for small row-size tables uses 2K, tablespace for large row-size tables uses 8K. Fall 2012 10 ITEC 450

11 M ODULE 4 D ATABASE T UNING Section 6 Storage Management 11 ITEC 450 Fall 2012

12 S PACE M ANAGEMENT As a DBA, you should track the following: Objects approaching an out of space condition Free space available – physical and logical perspective Segment or partition size – understand large or small tables Number of extents – less than hundreds Fragment usage information – performance concern Amount of reserved space that is currently unused – any waste situation Fall 2012 12 ITEC 450

13 D ATA P AGE L AYOUTS Three basic components exist in a data page (block in Oracle term) Page header – general page information, such as a page identifier, an identifier indicating to which table the page belongs, free space pointers Offset table – pointers to each data row on the data page Data rows – actual rows of user data (table or index) Row header – general row information, such as row length, information on variable-length data Offset tables – pointers to manage and control for variable- length fields Row data – actual data for the row Fall 2012 13 ITEC 450

14 RAID T ECHNOLOGY (W EEK 6) RAID is an acronym for Redundant Arrays of Inexpensive Disks. Please refer http://www.acnc.com/04_01_00.html RAID-0 – data striping: striped disk array without fault tolerance. RAID-1 – data mirroring: all data is written to two or more drives. RAID-5 – data stripping with distributed parity blocks, and it requires a minimum of 3 drives to implement (commonly 5 drives) RAID-10 – mirrored data stripping: very high reliability with high performance. Fall 2012 14 ITEC 450

15 M ODERN S TORAGE T ECHNOLOGY Storage Area Networks (SAN) – interconnected network of storage devices (dedicate connections between host and storage) Shared storage between multiple hosts High I/O performance Server and storage consolidation Used for database files, I/O intensive/high performance Network-Attached Storage (NAS) – storage that can be accessed directly from the network Shared storage between multiple hosts Simpler management due to reducing duplicate storage Application based storage access at file level Good for shared folders, program files Fall 2012 15 ITEC 450

16 M ODULE 4 D ATABASE T UNING Section 7 Data Movement and Distribution in Oracle 16 ITEC 450 Fall 2012

17 T HE LOAD U TILITY Oracle has SQL*Loader utility to perform bulk inserts of data into database tables. You can transform data before loading You can selectively load from the input file based on conditions You can load all or part of a table You can perform simultaneous data loads Two main steps to use SQL*Loader Prepare the data file – contains data in certain formats, field terminators Create a control file – defines how to map the data fields to a table and specifies if the data needs to be transformed Fall 2012 17 ITEC 450

18 D ATA P UMP E XPORT AND I MPORT Oracle offers the Data Pump technology for fast data movement between Oracle databases. Migrating databases from development to test or production Copying data between dev/test and production databases Transferring data between Oracle databases on different operating system platforms Backing up important tables or databases Reorganizing fragmented table data Extracting the DDL for tables and other objects Fall 2012 18 ITEC 450

19 D ATA P UMP E XPORT E XAMPLES Create a directory in Operating System for the dump directory C:\app\Administrator\admin\orcl450\dpdir1 Within Oracle, define and verify the directory for Data Pump SQL> create directory dpdir1 as C:\app\Administrator\admin\orcl450\dpdir1; SQL> select * from dba_directories where directory_name like 'DP%'; Export a table from command line (Run-> cmd) C:\> expdp system/password TABLES=hr.employees DIRECTORY=dpdir1 dumpfile=output01.dmp Export a schema C:\> expdp system/password SCHEMAS=hr DIRECTORY=dpdir1 dumpfile=output02.dmp You can also export a tablespace or full database Fall 2012 19 ITEC 450

20 D ATA P UMP I MPORT E XAMPLES You can import the information from the dump file generated from Export. Import to extract the DDL from the export dump file C:\> impdp system/password DIRECTORY=dpdir1 dumpfile=output01.dmp SQLFILE=emp01.sql Import a table to another schema C:\> impdp system/password DIRECTORY=dpdir1 dumpfile=output01.dmp REMAP_SCHEMA=hr:scott You can also import a schema, tablespace or full database Fall 2012 20 ITEC 450

21 D ATABASE R EFRESHING O PTIONS Backup and restore Backup the source database, and restore to the target location Commonly used for full database refresh Bath databases have similar physical structures Data pump Export the source database/schema/table, and then import to the target database Can be used for full, schema, or table level database refresh SQL*Loader Dump the source database tables to files in certain format, and then use SQL*Loader to upload the table data Often used for table refresh, different DMBS, or user ad- hoc files Fall 2012 21 ITEC 450

22 L OCK C ONFLICTS A share lock partially locks data where there is still partial access allowed to data by other sessions An exclusive lock completely prohibits changes to data, but still allows read access Locking of transactions and tables (using transactional control commands or LOCK TABLE) create locking situations manually Fall 2012 22 ITEC 450

23 H OW TO D ETECT L OCK C ONFLICTS Fall 2012 23 ITEC 450

24 H OW TO D ETECT L OCK C ONFLICTS Fall 2012 24 ITEC 450

25 H OW TO D ETECT L OCK C ONFLICTS Fall 2012 25 ITEC 450

26 W HAT IS A D EADLOCK ? Deadlock: one thing is waiting for another, but that other thing is also waiting for something A pending lock request cannot be serviced, because the lock required will never be released Can cause serious performance issues Most common cause is manual locking of data using LOCK TABLE or the FOR UPDATE clause Usually deadlocked sessions must be rolled back manually Ways of resolving a deadlock manually are: Rollback one of the deadlocked transactions Kill one of the sessions executing one of the deadlocked transactions Fall 2012 26 ITEC 450


Download ppt "Q UESTIONS ON B ACKUP AND R ECOVERY 1. What is the value of cold backup? 2. What's the difference between startup and startup mount? When we do "alter."

Similar presentations


Ads by Google