Slideshow:

Identifying a record (on disk)

There are 2 ways to represent the offset in a database address

Method 1: store offset as part of the record address

Method 2: store offset as an index in the offset table (in block header)

The block header contains a table of offset values !!

Record address contains an index in this offset table

Record deletion when using record offset

We must update all record references (hard)...

Record deletion when using record offset

Record deletion when using index

❮ ❯

Recall: how to identify/locate a record using its physical record address
- A record is identified/located by:
  Graphically:

2 way to represent the offset

There are 2 common ways to represent the offset of a record:

Using DatabaseAddress(Block X) + the (actual) offset in the block
Example: records Record 1 and Record 2 references the record A using block ID + offset:
Using DatabaseAddress(Block X) + the index in the offset table (= list of offsets in the block header)
Example: records Record 1 and Record 2 references the record A using block ID + index in offset table:

Advantage of using "DB-address + index in offset table" (case of trade-off)

Consider the deletion of a record when we use physical record address:

Suppose 2 records (Record 1 and Record 2) references a record using an (direct) offset:
We want to delete the record A
The deletion is done by:
Result:
The tombstone record is necessary because the records 1 and 2 (and there may be more) can not be easily found (so their references can be updated)

Consider the update procedure when using a record index:

Suppose 2 records (Record 1 and Record 2) references a record using an (indirect) index to an offset table:
We want to delete the record A
The deletion is done by:
Result:
We do not need to use the special tomb stone record (take up space)
We can re-use the all the space used by the deleted record
(Works especially well with fixed length fields)