Manjeet Singh.pptx

Parallel and Distributed Databases
1
By:- Manjeet Singh
Barr Code :-2140131
Group :- 220

Introduction
⚫What is a Centralized Database ?
-all the data is maintained at a single site and assumed that the processing of
individual transaction is essentially sequential.
2

PARALLEL DBMSs
WHY DO WE NEED THEM?
3
1
• Moreand More Data!
We havedatabases that hold a high amountof
data, in theorderof 1012 bytes:
10,000,000,000,000 bytes!
• Fasterand FasterAccess!
We havedataapplications that need toprocess
dataatvery high speeds:
10,000s transactionspersecond!
SINGLE-PROCESSOR DBMS AREN’T UP TO THE JOB!

5
Why Parallel Access To Data?
1 Terabyte
At 10 MB/s
1.2 days to scan
1 Terabyte
1,000 x parallel
1.5 minute to scan.
10 MB/s
Parallelism:
divide a big problem
into many smaller ones
to be solved in paralle

Parallel DB
⚫ Parallel database
5
system seeks to improve performance through
parallelization of various operations such as loading data ,building
indexes, and evaluating queries by using multiple CPUs and Disks in
Parallel.
⚫ Motivation for Parallel DB
⚫ Parallel machines are becoming quite common and affordable
⚫ Prices of microprocessors, memory and disks have dropped sharply
⚫ Databases are growing increasingly large
⚫ large volumes of transaction data are collected and stored for later
analysis.
⚫ multimedia objects like images are increasingly stored in databases

PARALLEL DBMSs
6
BENEFITS OF A PARALLEL DBMS
 Improves ResponseTime.
INTERQUERY PARALLELISM
It is possible to processa numberof transactions in
parallel with each other.
 ImprovesThroughput.
INTRAQUERYPARALLELISM
It is possible to process ‘sub-tasks’ of a transaction in
parallel with each other.

 Speed-Up
– Adding more resources results in proportionally less running time for a
fixed amount of data.
10 seconds to scan a DB of 10,000 records using 1 CPU
1 second to scan a DB of 10,000 records using 10 CPUs
PARALLEL DBMSs
HOW TO MEASURE THE BENEFITS
 Scale-Up
 If resources are increased in proportion to an increase in data/problem
size, the overall time should remain constant
– 1 second to scan a DB of 1,000 records using 1 CPU
1 second toscan a DB of 10,000 recordsusing 10 CPUs
7

Architectures for Parallel Databases
⚫ The basic idea behind Parallel DB is to carry out evaluation steps in
parallel whenever is possible.
⚫ There are many opportunities for parallelism in RDBMS.
⚫ 3 main architectures have been proposed for building parallel DBMSs.
1. Shared Memory
2. Shared Disk
3. Shared Nothing
8

Shared Memory
⚫ Advantages:
1. It is closer to conventional
machine , Easy to program
2. overhead is low.
3. OS services are leveraged to
utilize the additional CPUs.
⚫ Disadvantage:
1. It leads to bottleneck problem
2. Expensive to build
3. It is less sensitive to
partitioning
9

Shared Disk
⚫ Advantages:
1. Almost same
⚫ Disadvantages:
1. More interference
2. Increases N/W band width
3. Shared disk less sensitive to
partitioning
10

Shared Nothing
⚫ Advantages:
1. It provides linear scale up
&linear speed up
2. Shared nothing benefits from
"good" partitioning
3. Cheap to build
⚫ Disadvantage
1. Hard to program
2. Addition of new nodes
requires reorganizing
11

Sub-linear speed-up
Linear speed-up (ideal)
Number
of
transactions/second
1000/Sec
5 CPUs
2000/Sec
10 CPUs 16 CPUs
1600/Sec
2/12/201
Numberof CPUs
1. Parallel DB /D.S.Jagli 3
13
12
1. Parallel DB /D.S.Jagli
PARALLEL DBMSs
SPEED-UP

10 CPUs
2 GB Database
Number
of
transactions/second
Linear scale-up (ideal)
Sub-linear scale-up
1000/Sec
5 CPUs
1 GB Database
900/Sec
2/12/201
1. Parallel DB /D.S
N.Ja
ugli
mberof CPUs, Database size
3
14
13
PARALLEL DBMSs
SCALE-UP

PARALLEL QUERY EVALUATION
A relational query execution plan is graph/tree of
relational algebra operators (based on this operators can
execute in parallel)
1. 1. Parallel DB /D.S.Jagli
2/12/2013
15
14

Different Types of DBMS ||-ism
⚫ Parallel evaluation of a relational query in DBMS With shared –nothing
architecture
1. Inter-query parallelism
⚫ Multiple queries run on different sites
2. Intra-query parallelism
⚫ Parallel execution of single query run on different sites.
a) Intra-operator parallelism
a) get all machines working together to compute a given operation (scan, sort,
join).
b) Inter-operator parallelism
⚫ each operator may run concurrently on a different site (exploits
pipelining).
⚫ In order to evaluate different operators in parallel, we need to
evaluate each operator in query plan in Parallel.
1. 1. Parallel DB /D.S.Jagli
2/12/2013
15
15

Data Partitioning
2/12/2013
17
16
⚫ Types of Partitioning
1. Horizontal Partitioning: tuple of a relation are divided among
many disks such that each tuple resides on one disk.
⚫ It enables to exploit the I/O band width of disks by reading & writing
them in parallel.
relations from disk by
⚫ Reduce the time required to retrieve
partitioning the relations on multiple disks.
1. Range Partitioning
2. Hash Partitioning
3. Round Robin Partitioning
2. Vertical Partitioning

1.Range Partitioning
⚫ Tuples are sorted (conceptually), and n ranges are chosen for
the sort key values so that each range contains roughly the
same number of tuples;
⚫ tuples in range i are assigned to processor i.
⚫ Eg:
⚫ sailor _id 1-10 assigned to disk 1
⚫sailor _id 10-20 assigned to disk 2
⚫sailor _id 20-30 assigned to disk 3
⚫ range partitioning can lead to data skew; that is, partitions with widely
varying number of tuples across
17

2.Hash Partitioning
⚫ Ahash function is applied to selected fields of a tuple to determine its
processor.
⚫ Hash partitioning has the additional virtue that it keeps data evenly
distributed even if the data grows and shrinks over time.
18

3.Round Robin Partitioning
⚫ If there are n processors, the i th tuple is assigned to processor i mod n in
round-robin partitioning.
⚫ Round-robin partitioning is suitable for efficiently evaluating queries that
access the entire relation.
⚫ If only a subset of the tuples (e.g., those that satisfy the selection
condition age = 20) is required, hash partitioning and range partitioning
are better than round-robin partitioning
19

Range Hash Round Robin
A...E F...J K...N O...S T
...Z A...E F...J K...N O...
S
T...Z A...E F...J K...N O...
S
T...Z
Good for equijoins,
exact-match queries,
and range queries
20
Good for equijoins,
exact match queries
Good to spread load

Parallelizing Sequential Operator
Evaluation Code
1. An elegant software architecture for parallel DBMSs enables us to
readily parallelize existing code for sequentially evaluating a
relational operator.
2. The basic idea is to use parallel data streams.
3. Streams are merged as needed to provide the inputs for a relational
operator.
4. The output of an operator is split as needed to parallelize subsequent
processing.
5. A parallel evaluation plan consists of a dataflow network of
relational, merge, and split operators.
21

PARALLELIZING INDIVIDUAL
OPERATIONS
⚫ How various operations can be implemented in parallel in a shared-
nothing architecture?
⚫ Techniques
1. Bulk loading& scanning
2. Sorting
3. Joins
22

1.Bulk Loading and scanning
⚫ scanning a relation: Pages can be read in parallel while scanning a
relation, and the retrieved tuples can then be merged, if the relation is
partitioned across several disks.
⚫ bulk loading: if a relation has associated indexes, any sorting of data
entries required for building the indexes during bulk loading can also
be done in parallel.
2/12/2013
22
23

2.Parallel Sorting :
2/12/2013
25
1. Parallel DB /D.S.Jagli 24
⚫ Parallel sorting steps:
1. First redistribute all tuples in the relation using range partitioning.
2. Each processor then sorts the tuples assigned to it
3. The entire sorted relation can be retrieved by visiting the processors in
an order corresponding to the ranges assigned to them.
⚫ Problem: Data skew
⚫ Solution: “sample” the data at the outset to determine good
range partition points.
Aparticularly important application of parallel sorting is sorting the data
entries in tree-structured indexes.

TWO-PHASE COMMIT (2PC) - commit
1. Parallel DB /D.S.Jagli 2/12/2013
25

TWO-PHASE COMMIT (2PC) - ABORT
1. Parallel DB /D.S.Jagli 2/12/2013
26

Manjeet Singh.pptx

More Related Content

Similar to Manjeet Singh.pptx (20)

More from RAMCHANDRASHARMA7 (7)

Recently uploaded (20)

Manjeet Singh.pptx