- Database Address:
- Database Address =
an address used to identify
data (block/record)
stored on
disk
- Database Address =
an address used to identify
data (block/record)
stored on
disk
- How/when to use a
database address:
- When a record x need to
refer to
another record y,
we can store:
- the database address of y inside x
Example:
- When a record x need to
refer to
another record y,
we can store:
- There are 2 common techniques to
identify a
block/record stored on
disk:
- Physical datbase address
- Logical database address
- Physical Address
of a block:
- Physical Address of a
block =
physical address used to
identify a block
on a disk
- Components of a
physical block address:
A physical block address consists of:
- Host ID of the host (computer) that contains the disk
- Disk ID of the disk on the host
- Block number of the block on the disk
- Physical Address of a
block =
physical address used to
identify a block
on a disk
- Physical Address of a
record:
- Physical Address of a
record =
physical address used to
identify a record
on a disk
- Components of a
record physical address:
- Physical address of a block plus
- Offset of the record inside the block
- Physical Address of a
record =
physical address used to
identify a record
on a disk
- Logical (block/record) address:
- Each block/record is
assigned a unique
logical address
Each (valid) logical address is mapped to a physical address (of a block/record)
Example: map table
Logical Address Physical Address -------------------+------------------------------------ 101001.......10 | (HostID1, DiskID1, BlockID1) 010101.......01 | (HostID2, DiskID2, BlockID2) ... | ...
- Logical address =
an arbitrary string of
fixed length
bits
(Logical addresses can be generated using a counter --- we add 1 to the counter when we need a new value)
- Each block/record is
assigned a unique
logical address
- Data structure for
the map table:
- Sample
map table:
Logical Address Physical Address -------------------+------------------------------------ 101001.......10 | (HostID, DiskID, ..., BlockID) 010101.......01 | (HostID, DiskID, ..., BlockID) ...
- To speed up the
look up operation
(using a logical address to
find a physical address),
the map table is
organized
as:
- a hash table.
- The map table is stored on disk (in a file)
- Sample
map table:
- How to use the map table:
- When given a
logical database address,
we can find the
corresponding
physical database address
(where the block/record is
found physically) as
follows:
Explanation:
- Hash the logical database address
- Locate the entry in map table
- If entry is
valid:
return the (corresponding)
physical database address
If entry is not valid: return t invalid (not found)
- When given a
logical database address,
we can find the
corresponding
physical database address
(where the block/record is
found physically) as
follows:
- How to use
logical addresses:
- When a record x need to
refer to
another record y,
we now store:
- the logical database address of y inside x
Example:
- When a record x need to
refer to
another record y,
we now store:
- Problem with
referencing another record
using a physical
(database) address:
- Example:
2 records X1 and X2
are referencing the
record Y
using
physical database addresses:
- Problem:
If record Y is
moved (stored) to a
different block on
the disk:
- We must update
multiple (=all)
(physical) database addresses:
- We must update
multiple (=all)
(physical) database addresses:
- Example:
2 records X1 and X2
are referencing the
record Y
using
physical database addresses:
- Better way:
store
logical (database) addresses as
reference inside
records:
- Example: 2 records X1 and X2
references the
record Y
using
logical database addresses:
- If we move the
record Y, we
only need to:
- Update the
(one) physical address in the
map table:
- Update the
(one) physical address in the
map table:
- Example: 2 records X1 and X2
references the
record Y
using
logical database addresses:
- Consider the records
stored in a block:
Clearly:
- Record address = address of the block that contains the record + an offset
- There are 2 common ways
to identify
a record:
- (DatabaseAddress(Block X)
+ the
(absolute) offset in block)
Example:
- record addr 0: (DatabaseAddress(Block X), 456 (offset in bytes))
- record addr 1:
(DatabaseAddress(Block X),
234 (offset in bytes))
- (DatabaseAddress(Block X) +
the index
in an offset table
(stored in the block header)
Example:
- record addr 1: (DatabaseAddress(Block X), 1)
- record addr 2: (DatabaseAddress(Block X), 0)
Inside the block header, we find an offset table:
Offset table: 234 456 ....Graphically:
- (DatabaseAddress(Block X)
+ the
(absolute) offset in block)
- Recall:
Database addresses
- Database addresses are part of a record and is used to reference another record
- Example usage 1: reference to
your supervisor
- Employee "John"
has the supervisor
employee Frank
- Employee "John"
has the supervisor
employee Frank
- Example usage 2: index files
-
Index files are
constructed to
speed up access
to
the data file
(We will study index files in the next chapter)
- Index files need to use
database addresses to
locate the records:
-
Index files are
constructed to
speed up access
to
the data file