MBA 664Database Management Systems Dave Salisbury email@example.com (email) http://www.davesalisbury.com/ (web site)
Physical Database Design • The purpose of the physical design process is to translate the logical description of the data into technical specifications for storing and retrieving data • Goal: create a design that will provide adequate performance and insure database integrity, security, and recoverability • Decisions made in this phase have a major impact on data accessibility, response times,security, and user friendliness.
Inputs • Normalized relations • Volume estimates • Attribute definitions • Response time expectations • Data security needs • Backup/recovery needs • Integrity expectations • DBMS technology used Physical Design Process Decisions • Attribute data types • Physical record descriptions (doesn’t always match logical design) • File organizations • Indexes and database architectures • Query optimization Leads to
Determining volume and usage • Data volume statistics represent the size of the business • calculated assuming business growth over a period of several years • Usage is estimated from the timing of events, transaction volumes, and reporting and query activity. • Less precise than volume statistics
Figure 6.1 - Composite usage map (Pine Valley Furniture Company)
Figure 6.1 - Composite usage map (Pine Valley Furniture Company) Data volumes
Figure 6.1 - Composite usage map (Pine Valley Furniture Company) Access Frequencies (per hour)
Figure 6.1 - Composite usage map (Pine Valley Furniture Company) Usage analysis: 200 purchased parts accessed per hour 80 quotations accessed from these 200 purchased part accesses 70 suppliers accessed from these 80 quotation accesses
Figure 6.1 - Composite usage map (Pine Valley Furniture Company) Usage analysis: 75 suppliers accessed per hour 40 quotations accessed from these 75 supplier accesses 40 purchased parts accessed from these 40 quotation accesses
Physical Design Decisions • Specify the data type for each attribute from the logical data model • minimize storage space and maximize integrity • Specify physical records by grouping attributes from the logical data model • Specify the file organization technique to use for physical storage of data records • Specify indexes to optimize data retrieval • Specify query optimization strategies
Designing Fields • Field: smallest unit of data in database • Field design • Choosing data type • Coding, compression, encryption • Controlling data integrity
Choosing Data Types • CHAR – fixed-length character • VARCHAR2 – variable-length character (memo) • LONG – large number • NUMBER – positive/negative number • DATE – actual date • BLOB – binary large object (good for graphics, sound clips, etc.)
Data Format • Data type selection goals • minimize storage • represent all possible values • eliminate illegal values • improve integrity • support manipulation • Note: these have different relative importance
Data format decisions (coding) • E.G., C(OAK), B(MAPLE) , etc • Implement by creating a look-up table • There is a trade-off in that you must create and store a second table and you must access this table to look up the code value • Consider using when a field has a limited number of possible values, each of which occupies a relatively large amount of space, and the number of records is large and/or the number of record accesses is small
Figure 6.2 Example code-look-up table (Pine Valley Furniture Company) Code saves space, but costs an additional lookup to obtain actual value.
Data integrity controls Default value Range control Null value control Referential integrity Missing data substitute an estimate report missing data sensitivity testing Triggers can be used to perform these operations Data Format decisions (integrity)
For example... • Suppose you were designing the age field in a student record at your university. What decisions would you make about: • data type • integrity (range, default, null) • How might your decision vary by other characteristics about the student such as degree sought?
Physical Records • Physical Record: A group of fields stored in adjacent memory locations and retrieved together as a unit • Page: The amount of data read or written in one I/O operation • Blocking Factor: The number of physical records per page
Denormalization • Transforming normalized relations into unnormalized physical record specifications • Benefits: • Can improve performance (speed) be reducing number of table lookups (i.e reduce number of necessary join queries) • Costs (due to data duplication) • Wasted storage space • Data integrity/consistency threats • Common denormalization opportunities • One-to-one relationship (Fig 6.3) • Many-to-many relationship with attributes (Fig. 6.4) • Reference data (1:N relationship where 1-side has data not used in any other relationship) (Fig. 6.5)
Fig 6.5 – A possible denormalization situation: reference data Extra table access required Data duplication
Consider the following normalized relations • STORE(Store_Id, Region, Manager_Id, Square_Feet) • EMPLOYEE(Emp_Id, Store_Id, Name, Address) • DEPARTMENT(Dept#, Store_ID, Manager_Id, Sales_Goal) • SCHEDULE(Dept#, Emp_Id, Date, hours) What opportunities might exist for denormalization?
Partitioning • Horizontal Partitioning: Distributing the rows of a table into several separate files • Useful for situations where different users need access to different rows • Three types: Key Range Partitioning, Hash Partitioning, or Composite Partitioning • Vertical Partitioning: Distributing the columns of a table into several separate files • Useful for situations where different users need access to different columns • The primary key must be repeated in each file • Combinations of Horizontal and Vertical Partitions often correspond with User Schemas (user views)
Partitioning • Advantages of Partitioning: • Records used together are grouped together • Each partition can be optimized for performance • Security, recovery • Partitions stored on different disks: contention • Take advantage of parallel processing capability • Disadvantages of Partitioning: • Slow retrievals across partitions • Complexity
Data Replication • Purposely storing the same data in multiple locations of the database • Improves performance by allowing multiple users to access the same data at the same time with minimum contention • Sacrifices data integrity due to data duplication • Best for data that is not updated often
Designing Physical Files • Physical File: • A named portion of secondary memory allocated for the purpose of storing physical records • Constructs to link two pieces of data: • Sequential storage. • Pointers. • File Organization: • How the files are arranged on the disk. • Access Method: • How the data can be retrieved based on the file organization.
Figure 6-7 (a) Sequential file organization 1 2 If sorted – every insert or delete requires resort Records of the file are stored in sequence by the primary key field values. If not sorted Average time to find desired record = n/2. n
Sequential Retrieval • Consider a file of 10,000 records each occupying 1 page • Queries that require processing all records will require 10,000 accesses • e.g., Find all items of type 'E' • Many disk accesses are wasted if few records meet the condition • However, very effective if most or all records will be accessed (e.g., payroll)
Indexed File Organizations • Index – a separate table that contains organization of records for quick retrieval – like an index in a book. • Primary keys are automatically indexed • Oracle has a CREATE INDEX operation, and MS ACCESS allows indexes to be created for most field types • Indexing approaches: • B-tree index, Fig. 6-7b • Bitmap index, Fig. 6-8 • Hash Index, Fig. 6-7c • Join Index, Fig 6-9
Fig. 6-7b – B-tree index uses a tree search Average time to find desired record = depth of the tree Leaves of the tree are all at same level consistent access time
Hashed File Organization • Hashing Algorithm: Converts a primary key value into a record address • Division-remainder method is common hashing algorithm
Hashing • A technique for reducing disk accesses for direct access • Avoids an index • Number of accesses per record can be close to one • The hash field is converted to a hash address by a hash function
Shortcomings of Hashing • Different hash fields may convert to the same hash address • these are called Synonyms • store the colliding record in an overflow area • Long synonym chains degrade performance • There can be only one hash field per record • The file can no longer be processed sequentially • More collisions between synonyms leads to reduced access speed
Fig 6-7c Hashed file or index organization Hash algorithm Usually uses division-remainder to determine record position. Records with same position are grouped in lists.
Bitmap saves on space requirements Rows - possible values of the attribute Columns - table rows Bit indicates whether the attribute of a row has the values Fig 6-8 Bitmap index index organization
Clustering Files • In some relational DBMSs, related records from different tables can be stored together in the same disk area • Useful for improving performance of join operations • Primary key records of the main table are stored adjacent to associated foreign key records of the dependent table • e.g. Oracle has a CREATE CLUSTER command
Indexing • An index is a table file that is used to determine the location of rows in another file that satisfy some condition
Querying with an Index • Read the index into memory • Search the index to find records meeting the condition • Access only those records containing required data • Disk accesses are substantially reduced when the query involves few records
Maintaining an Index • Adding a record requires at least two disk accesses: • Update the file • Update the index • Trade-off: • Faster queries • Slower maintenance (additions, deletions, and updates of records) • Thus, more static databases benefit more overall
Rules for Using Indexes 1. Use on larger tables 2. Index the primary key of each table 3. Index search fields (fields frequently in WHERE clause) 4. Fields in SQL ORDER BY and GROUP BY commands 5. When there are >100 values but not when there are <30 values
Rules for Using Indexes 6. DBMS may have limit on number of indexes per table and number of bytes per indexed field(s) 7. Null values will not be referenced from an index 8. Use indexes heavily for non-volatile databases; limit the use of indexes for volatile databases Why? Because modifications (e.g. inserts, deletes) require updates to occur in index files
Rules for Adding Derived Columns • Use when aggregate values are regularly retrieved. • Use when aggregate values are costly to calculate. • Permit updating only of source data. • Create triggers to cascade changes from source data.
One Other Rule of Thumbfor Increasing Performance • Consider contriving a shorter field or selecting another candidate key to substitute for a long, multi-field primary key (and all associated foreign keys)
RAID • Redundant Arrays of Inexpensive Disks • Exploits economies of scale of disk manufacturing for the computer market • Can give greater security • Increases fault tolerance of systems • Not a replacement for regular backup
RAID • The operating system sees a set of physical drives as one logical drive • Data are distributed across physical drives • All levels, except 0, have data redundancy or error-correction features • Parity codes or redundant data are used for data recovery
Mirroring • Write • Identical copies of file are written to each drive in array • Read • Alternate pages are read simultaneously from each drive • Pages put together in memory • Access time is reduced by approximately the number of disks in the array • Read error • Read required page from another drive • Tradeoffs • Provides data security • Reduces access time • Uses more disk space
Mirroring Complete Data Set Complete Data Set No parity
Striping • Three drive model • Write • Half of file to first drive • Half of file to second drive • Parity bit to third drive • Read • Portions from each drive are put together in memory • Read error • Lost bits are reconstructed from third drive’s parity data • Tradeoffs • Provides data security • Uses less storage space than mirroring • Not as fast as mirroring
Striping One-Half Data Set One-Half Data Set Parity Codes