Storage Management - Lecture 8 - Introduction to Databases (1007156ANR)

This lecture is part of an Introduction to Databases course given at the Vrije Universiteit Brussel.

1. 2 December 2005
Introduction to Databases
Storage Management
Prof. Beat Signer
Department of Computer Science
Vrije Universiteit Brussel


http://www.beatsigner.com

2. Beat Signer - Department of Computer Science - [email protected]

2 April 19, 2013

Context of Today's Lecture

Access
Methods
System
Buffers
Authorisation
Control
Integrity
Checker

Command
Processor
Program
Object Code
DDL
Compiler
File
Manager
Buffer
Manager
Recovery
Manager
Scheduler
Query
Optimiser
Transaction
Manager
Query
Compiler
Queries
Catalogue
Manager
DML
Preprocessor
Database
Schema
Application
Programs
Database
Manager
Data
Manager
DBMS
Programmers Users DB Admins
Based on 'Components of a DBMS', Database Systems,
T. Connolly and C. Begg, Addison-Wesley 2010
Data, Indices and
System Catalogue

3. Beat Signer - Department of Computer Science - [email protected]

3 April 19, 2013

Storage Device Hierarchy
Storage devices vary in
data capacity
access speed
cost per byte
Devices with fastest
access time have
highest costs and
smallest capacity
Cache
Main Memory
Flash Memory
Magnetic Disk
Optical Disk
Magnetic Tapes

4. Beat Signer - Department of Computer Science - [email protected]

4 April 19, 2013

Cache
On-board cache on the same chip as the microprocessor
level 1 (L1) cache
temporary storage of instructions and data
typical size of ~64 kB
Extra cache levels located on separate chips
e.g. level 2 (L2) cache
typical size of ~1 MB
Data items in the cache are copies of values in main
memory locations
If data in the cache has been updated, changes must be
reflected in the corresponding memory locations

5. Beat Signer - Department of Computer Science - [email protected]

5 April 19, 2013

Main Memory
Main memory can be several gigabytes large
Normally too small and too expensive for storing the
entire database
content is lost during power failure or crash (volatile memory)
in-memory databases (IMDB) primarily rely on main memory
-note that IMDBs lack durability (D of the ACID properties)
IMDB size limited by the maximal addressable memory space
-e.g. maximal 4 GB for 32-bit address space
Random access memory (RAM)
time to access data is more or less independent of its location
(different from magnetic tapes)
Typical access time of ~10 nanoseconds (10
-8
seconds)

6. Beat Signer - Department of Computer Science - [email protected]

6 April 19, 2013

Secondary Storage (Hard Disk)
Essentially random access
Files are moved between a hard disk and main memory
(disk I/O) by the operating system (OS) or the DBMS
the transfer units are blocks
tendency for larger block sizes
Parts of the main memory are used to buffer blocks
the buffer manager of the DBMS manages the loading and
unloading of blocks for specific DBMS operations
Typical block I/O time (seek time) ~10 milliseconds
1'000'000 times slower than main memory access
Capacity of several hundred gigabytes and a system can
use many disk units

7. Beat Signer - Department of Computer Science - [email protected]

7 April 19, 2013

Hard Disk
A hard disk contains one
or more platters and one or
more heads
The platters were originally
addressed in terms of cylinders,
heads and sectors (block)
cylinder-head-sector (CHS) scheme
max of 1024 cylinders, 16 heads
and 63 sectors
Current hard disks offer
logical block addressing (LBA)
hides the physical disk geometry

8. Beat Signer - Department of Computer Science - [email protected]

8 April 19, 2013

Solid-State Drives (SSD)
Storage device that uses solid-state memory
(flash memory) to persistently store data
Offers a hard disk interface with a storage capacity of up
to a few hundred gigabytes
Typical block I/O time (seek time) ~0.1 milliseconds
SSDs might help to reduce the gap between primary and
secondary storage in DBMS systems
Currently there are still some limitations of SSDs
the limited number of SSD write operations before failure can be a
problem for DBs with a lot of update operations
write operations are often still much slower than read operations

9. Beat Signer - Department of Computer Science - [email protected]

9 April 19, 2013

Tertiary Storage
No random access
access time depends on
data location
Different devices
tapes
optical disk jukeboxes
-racks of CD-ROMs (read only)
tape silos
-room-sized devices holding
racks of tapes operated by
tape robots
-e.g. StorageTek PowderHorn
with up to 28.8 petabytes

10. Beat Signer - Department of Computer Science - [email protected]

10 April 19, 2013

Models of Computation
RAM model of computation
assumes that all data is held in main memory
DBMS model of computation
assumes that data does not fit into main memory
efficient algorithms must take into account secondary and even
tertiary storage
best algorithms for processing large amounts of data often differ
from those for the RAM model of computation
minimising disk accesses plays a major role
-I/O model of computation
I/O model of computation
the time to move a block between disk and memory is much
higher than the time for the corresponding computation

11. Beat Signer - Department of Computer Science - [email protected]

11 April 19, 2013

Accelerating Secondary Storage Access
Various possible strategies to improve secondary
storage access
placement of blocks that are often accessed together on the same
disk cylinder
distribute data across multiple disks to profit from parallel disk
accesses (e.g. RAID)
mirroring of data
use of disk scheduling algorithms in OS, DBMS or disk controller
to determine order of requested block read/writes
-e.g. elevator algorithm
prefetching of disk blocks
efficient caching
-main memory
-disk controllers

12. Beat Signer - Department of Computer Science - [email protected]

12 April 19, 2013

Redundant Array of Independent Disks
The redundant array of independent disks (RAID)
organisation technique provides a single disk view for a
number (array) of disks
divide and replicate data across multiple hard disks
introduced in 1987 by D.A. Patterson, G.A. Gibson and R. Katz
The main goals of a RAID solution are
higher capacity by grouping multiple disks
-originally a RAID was also a cheaper alternative to expensive large disks
•original name: Redundant Array of Inexpensive Disks
higher performance due to parallel disk access
-multiple parallel read/write operations
increased reliability since data might be stored redundantly
-data can be restored if a disk fails

13. Beat Signer - Department of Computer Science - [email protected]

13 April 19, 2013

RAID ...
There are three main concepts in RAID systems
identical data is written to more than one disk (mirroring)
data is split accross multiple disks (striping)
redundant parity data is stored on separated disks and used to
detect and fix problems (error correction)

14. Beat Signer - Department of Computer Science - [email protected]

14 April 19, 2013

RAID Reliability
The mean time between failures (MTBF) is the average
time until a disk failure occurs
e.g. a hard disk might have a MTBF of 200'000 hours (22.8 years)
-note that the MTBF decreases as disks get older
If a DBMS uses an array of disks, then the overall
system's MTBF can be much lower
e.g. the MTBF for a disk array of 100 of the disks mentioned
above is 200'000 hours/100 = 2'000 hours (83 days)
By storing information redundantly, data can be restored
in the case of a disk failure

15. Beat Signer - Department of Computer Science - [email protected]

15 April 19, 2013

RAID Reliability ...
The mean time to data loss (MTTDL) depends on the
MTBF and the mean time to repair
if we mirror the information on two disks with a MTBF of 200'000
hours and a mean time to repair of 10 hours then the MTTDL is
200'000
2
/(2*10) hours = 228'000 years
of course in reality it is more likely that an error occurs on multiple
disks around the same time
-drives have the same age
-power failure, earthquake, fire, ...

16. Beat Signer - Department of Computer Science - [email protected]

16 April 19, 2013

RAID Levels
The different RAID levels offer different
cost-performance trade-offs
RAID 0
block level striping without any redundancy
RAID 1
mirroring without striping
RAID 2
bit level striping
multiple parity disks
RAID 3
byte level striping
one parity disk
[http://en.wikipedia.org/wiki/RAID]

17. Beat Signer - Department of Computer Science - [email protected]

17 April 19, 2013

RAID Levels ...
RAID 4
block level striping
one parity disk
similar to RAID 3
RAID 5
block level striping with distributed parity
no dedicted parity disk
RAID 6
block level striping with dual
distributed parity
no dedicted parity disk
similar to RAID 5

18. Beat Signer - Department of Computer Science - [email protected]

18 April 19, 2013

Data Representation
A DBMS has to define how the elements of its data model
(e.g. relational model) are mapped to secondary storage
a field contains a fixed- or variable-length sequence of bytes and
represents an attribute
a record contains a fixed- or variable-length sequence of fields and
represents a tuple
records are stored in fixed-length physical block storage units
representing a set of tuples
-the blocks also represent the units of data transfer
a file contains a collection of blocks and represents a relation
A database is finally mapped to a number of files
managed by the underlying operating system
index structures are stored in separate files

19. Beat Signer - Department of Computer Science - [email protected]

19 April 19, 2013

Relational Model Representation
A number of issues have to be addressed when
mapping the basic elements of the relational model
to secondary storage
how to map the SQL datatypes to fields?
how to represent tuples as records?
how to represent records in blocks?
how to represent a relation as a collection of blocks?
how to deal with record sizes that do not fit into blocks?
how to deal with variable-length records?
how to deal with schema updates and growing record lengths?
...

20. Beat Signer - Department of Computer Science - [email protected]

20 April 19, 2013

Representation of SQL Datatypes
Fixed-length character string (CHAR(n))
represented as a field which is an array of n bytes
strings that are shorter than n bytes are filled up with a special
"pad" character
Variable-length character string (VARCHAR(n))
two common representations (non-fixed length version later)
length plus content
-allocate an array of n + 1 bytes
-the first byte represents the length of the string (8-bit integer) followed by the
string content
-limited to a maximal string length of 255 characters
null-terminated string
-allocate an array of n + 1 bytes
-terminate the string with a special null character (like in C)

21. Beat Signer - Department of Computer Science - [email protected]

21 April 19, 2013

Representation of SQL Datatypes ...
Dates (DATE)
fixed-length character string
Time (TIME(n))
the precision n leads to strings of variable length and two possible
representations
fixed-precision
-limit the precision to a fixed value and store as VARCHAR(m)
true-variable length
-store the time as true variable length value
Bits (BIT(n))
bit values of size n can be packed into single bytes
packing of multiple bit values into a single byte is not recommended
-makes the retrieval and updating of a value more complex and error-prone

22. Beat Signer - Department of Computer Science - [email protected]

22 April 19, 2013

Storage Access
A part of the system's main memory is used as a buffer
to store copies of disk blocks
The buffer manager is responsible to move data from
secondary disk storage into memory
the number of block transfers between disk and memory should
be minimised
as many blocks a possible should be kept in memory
The buffer manager is called by the DMBS every time a
disk block has to be accessed
the buffer manager has to check whether the block is already
allocated in the buffer (main memory)

23. Beat Signer - Department of Computer Science - [email protected]

23 April 19, 2013

Buffer Manager
If the requested block is already in the buffer, the buffer
manager returns the corresponding address
If the block is not yet in the buffer, the buffer manager
performs the following steps
allocate buffer space
-if no space is available, remove an existing block from the buffer
(based on a buffer replacement strategy) and write it back to the disk if it has
been modified since it was last fetched/written to disk
read the block from the disk, add it to the buffer and return the
corresponding memory address
Note the similarities to a virtual memory manager

24. Beat Signer - Department of Computer Science - [email protected]

24 April 19, 2013

Buffer Replacement Strategies
Most operating systems use a least recently used (LRU)
strategy where the block that was least recently used is
moved back from memory to disk
use past access pattern to predict future block access
A DBMS is able to predict future access patterns more
accurately than an operating system
a request to the DBMS involves multiple steps and the DBMS
might be able to determine which blocks will be needed by
analysing the different steps of the operation
note that LRU might not always be the best replacement strategy
for a DBMS

25. Beat Signer - Department of Computer Science - [email protected]

25 April 19, 2013

Buffer Replacement Strategies ...
Let us have a look at the procedure to compute the
following natural join query: order ⋈ customer
note that we will see more efficient solutions for this problem
when discussing query optimisation
for each tuple o of order {
for each tuple c of customer {
if o.customerID = c.customerID {
create a new tuple r with:
r.customerID := c.customerID
r.name := c.name
...
r.orderID := o.orderID
...
add tuple r to the result set of the join operation
}
}
}

26. Beat Signer - Department of Computer Science - [email protected]

26 April 19, 2013

Buffer Replacement Strategies ...
We further assume that the two relations order and
customer are stored in separate files
From the pseudocode we can see that
once an order tuple has been processed, it is not needed
anymore
-if a whole block of order tuples has been processed, that block is no longer
required in memory (but an LRU strategy might keep it)
-as soon as the last tuple of an order block has been processed, the buffer
manager should free the memory space  toss-immediate strategy
once a customer tuple has been processed, it is not accessed
again until all the other customer tuples have been accessed
-when the processing of a customer block has been finished, the least recently
used customer block will be requested next
-we should replace the block that has been most recently used (MRU)

27. Beat Signer - Department of Computer Science - [email protected]

27 April 19, 2013

Buffer Replacement Strategies ...
A memory block can be marked to indicate that this block
is not allowed to be written back to disk (pinned block)
note that if we want to use an MRU strategy for the inner loop of
the previous example, the block has to be pinned
-the block has to be unpinned after the last tuple in the block has be processed
the pinning of blocks provides some control to restrict the time
when blocks can be written back to disk
-important for crash recovery
-blocks that are currently updated should not be written to disk
The prefetching of blocks might be used to further
increase the performance of the overall system
e.g. for serial scans (relation scans)

28. Beat Signer - Department of Computer Science - [email protected]

28 April 19, 2013

Buffer Replacement Strategies ...
The buffer manager can also use statistical information
about the probability that a request will reference a
particular relation (and its related blocks)
the system catalogue (data dictionary) with its metadata is one of
the most frequently accessed parts of the database
-if possible, system catalogue blocks should always be in the buffer
index files might be accessed more often than the corresponding
files themselves
-do not remove index files from the buffer if not necessary
the crash recovery manager can also provide constraints for the
buffer manager
-the recovery manager might demand that other blocks have to be written first
(force-output) before a specific block can be written to disk

29. Beat Signer - Department of Computer Science - [email protected]

29 April 19, 2013

System Catalogue / Data Dictionary
Stores metadata about the database
names of the relations
names, domain and lengths of the attributes of each relation
names of views
names of indices
-name of relation that is indexed
-name of attributes
-type of index
integrity constraints
users and their authorisations
statistical data
-number of tuples in relation, storage method, ...
...

30. Beat Signer - Department of Computer Science - [email protected]

30 April 19, 2013

File Organisation
A file is a logically organised as a sequence of records
each record contains a sequence of fields
name, datatype and offset of record fields are defined by
the schema
record types (schema) might change over time
The records are mapped to disk blocks
the block size is fixed and defined by the physical properties of
the disk and the operating system
the record size might vary for different relations and even
between tuples of the same relation (variable field size)
There are different possible mappings of records to files
use multiple files and only store fixed-length records in each file
store variable-length records in a file

31. Beat Signer - Department of Computer Science - [email protected]

31 April 19, 2013

Fixed-Length Records




If we assume that an integer requires 2 bytes and
characters are represented by one byte, then the
customer record is 64 bytes long

type customer = record
cID int;
name varchar(30)
street varchar(30)
end

cID name street ... ...
0 2 33 64
Block

32. Beat Signer - Department of Computer Science - [email protected]

32 April 19, 2013

Fixed-Length Records ...


It might be necessary to ensure that data elements begin
at an offset that is a multiple of 4 (8 for 64-bit processors)
the first byte of a block loaded from disk is placed at a memory
address that is a multiple of 4
we have to ensure that we have the appropriate offsets
(e.g. dividable by 4)
cID name street ... ...
0 4 36 68
Block

33. Beat Signer - Department of Computer Science - [email protected]

33 April 19, 2013

Fixed-Length Records ...


Often a record header is added to each record for
managing metadata about
the record schema (pointer s to the DBMS schema information)
timestamp t about the last access or modification time
the length l of the record
-could be computed from the schema but the information is convenient if we
want to quickly access the next record without having to consult the schema
...


0 12 48 80
cID name street ... ... s t l
16
Block

34. Beat Signer - Department of Computer Science - [email protected]

34 April 19, 2013

Fixed-Length Records in Blocks/Files




Problems with this fixed length representation
after a record has been deleted, its space has to be filled with
another record
-could move all records after the deleted one but that is too expensive
-can move the last record to the deleted record's position but also that might
require an additional block access
if the block size is not a multiple of the record size, some records
will cross block boundaries and we need two block accesses to
read/write such a record
1 Max Frisch Bahnhofstrasse 7 h
2 Eddy Merckx Pleinlaan 25 h
5 Claude Debussy 12 Rue Louise h
53 Albert Einstein Bergstrasse 18 h
8 Max Frisch ETH Zentrum h
record 0
record 1
record 2
record 3
record 4

35. Beat Signer - Department of Computer Science - [email protected]

35 April 19, 2013

Fixed-Length Records in Blocks/Files ...
Since insert operations tend to be more frequent that
delete operations, it might be acceptable to leave the
space of the deleted record open until a new record is
inserted
we cannot just add an additional boolean flag ("free") to the
record since it will be hard to find the free records
allocate a certain amount of bytes for a file header containing
metadata about the file
The block/file header contains a pointer (address) to the
first deleted record
each deleted record contains a pointer (address) to the next
deleted record
the linked list of deleted records is called a free list

36. Beat Signer - Department of Computer Science - [email protected]

36 April 19, 2013

record 0
record 1
record 2
record 3
record 4
header
Fixed-Length Records in Blocks/Files ...





To insert a new record, the first free record pointed to by
the header is used and the address in the header is
updated to the free record that the used record was
pointing to
to save some space, the pointers of the free list can also be
stored in the unused space of deleted records (no additional field)

1 Max Frisch Bahnhofstrasse 7 h
5 Claude Debussy 12 Rue Louise h
8 Max Frisch ETH Zentrum h

37. Beat Signer - Department of Computer Science - [email protected]

37 April 19, 2013

Address Space
There are several ways how the database address
space (blocks and block offsets) can be represented
physical addresses consisting of byte strings (up to 16 bytes)
that address
-host
-storage device identifier (e.g. hard disk ID)
-cylinder number of the disk
-track within the cylinder (for multi-surface disks)
-block within the track
-potential offset of record within the block
logical addresses consisting of an arbitrary string of length n

38. Beat Signer - Department of Computer Science - [email protected]

38 April 19, 2013

Address Space Mapping




A map table is stored at a known disk location and
provides a mapping between the logical and physical
address spaces
introduces some indirection since the map table has to be
consulted to get the physical address
flexibility to rearrange records within blocks or move them to other
blocks without affecting the record's logical address
different combinations of logical and physical addresses are
possible (structured address schemes)
... ...
logical physical logical
address
physical
address
map table

39. Beat Signer - Department of Computer Science - [email protected]

39 April 19, 2013

Variable-Length Data
Records of the same type may have different lengths
We may want to represent
record fields with varying size (e.g. VARCHAR(n))
large fields (e.g. images)
...
We need an alternative data representation to deal with
these requirements

40. Beat Signer - Department of Computer Science - [email protected]

40 April 19, 2013

Variable-Length Record Fields



Scheme for records with variable-length fields
put all fixed-length fields first (e.g. cID)
add the length of the record to the record header
add the offsets of the variable-length fields to the record header
Note that if the order of the variable-length fields is
always the same, we do not have to store an offset for
the first variable-length field (e.g. name)
cID
name
street

record length

41. Beat Signer - Department of Computer Science - [email protected]

41 April 19, 2013

Variable-Length Records



There are different reasons why we might have to use
variable-length records
to store records that have at least one field with a variable length
to store different record types in a single block/file
Structured address scheme (slotted-page structure)
address of a record consists of the block address in combination
with an offset table index
records can be moved around
record
3 record
2 record
1 ... free ...
offset table

42. Beat Signer - Department of Computer Science - [email protected]

42 April 19, 2013

Large Records



Sometimes we have to deal with values that do not fit
into a single block (e.g. audio or movie clips)
a record that is split across two or more blocks is called
a spanned record
spanned records can also be used to pack blocks more efficiently
Extra header information
each record header carries a bit to indicate if it is a fragment
-fragments have some more bits; telling whether first or last fragment of record
potential pointers to previous and next fragment

block header
record header
record
2b record
3 record
2a record
1
block 1 block 2

43. Beat Signer - Department of Computer Science - [email protected]

43 April 19, 2013

Storage of Binary Large Objects (BLOBS)
BLOB is stored as a sequence of blocks
often blocks allocated successively on a disk cylinder
BLOB might be striped across multiple disks for more
efficient retrieval
BLOB field might not be automatically fetched into
memory
user has to explicitly load parts of the BLOB
possibly index structures to retrieve parts of a BLOB

44. Beat Signer - Department of Computer Science - [email protected]

44 April 19, 2013

Insertion of Records




If the records are not kept in a particular order, we can
just find a block with some empty space or create a new
block if there is no such space
If the record has to be inserted in a particular order, but
there is no space in the block, there are two alternatives
find space in a nearby block and rearrange some records
create an overflow block and link it from the header of the original
block
-note that an overflow block might point to another overflow block and so on

record
3 record
2 record
1 ... free ...
offset table

45. Beat Signer - Department of Computer Science - [email protected]

45 April 19, 2013

Deletion of Records




If we use an offset table, we may compact the free space
in the block (slide around the records)
If the records cannot be moved, we might have a free list
in the header
We might also be able to remove an overflow block after
a delete operation
record
3 record
2 record
1 ... free ...
offset table

46. Beat Signer - Department of Computer Science - [email protected]

46 April 19, 2013

Update of Records




If we have to update a fixed-length record there is no
problem since we will still use the same space
If the updated record is larger than the original version,
then we might have to create more space
same options as discussed for insert operation
If the updated record is smaller, then we may compact
some free space or remove overflow blocks
similar to delete operation
record
3 record
2 record
1 ... free ...
offset table

47. Beat Signer - Department of Computer Science - [email protected]

47 April 19, 2013

Homework
Study the following chapter of the
Database System Concepts book
chapter 10
-sections 10.1-10.9
-Storage and File Structure

48. Beat Signer - Department of Computer Science - [email protected]

48 April 19, 2013

Exercise 8
Structured Query Language (SQL)
PostgreSQL

49. Beat Signer - Department of Computer Science - [email protected]

49 April 19, 2013

References
H. Garcia-Molina, J.D. Ullman and J. Widom,
Database Systems – The Complete Book,
Prentice Hall, 2002
A. Silberschatz, H. Korth and S. Sudarshan, Database
System Concepts (Sixth Edition), McGraw-Hill, 2010

50. 2 December 2005
Next Lecture
Access Methods