TÀI LIỆU - Cao Học Khóa 8 - ĐH CNTT 5. a NoSql

TÀI LIỆU - Cao Học Khóa 8 - ĐH CNTT 5. a NoSql tài liệu, giáo án, bài giảng , luận văn, luận án, đồ án, bài tập lớn về t...

Trang 1

NoSQL Databases – Amir H. Payberah

Trang 3

NoSQL Databases – Amir H. Payberah 3 April 26, 2012

SQL is Good

● Relational Databases Management Systems (RDMBSs) – mainstay of business

Trang 4

Trang 5

SQL is Good

● SQL is good,

Trang 6

Trang 7

Trang 8

The Past and the Moment

http://www.couchbase.com/sites/default/files/uploads/all/whitepapers/NoSQLWhitepaper.pdf

Trang 9

Trang 10

Trang 11

Let's Scale RDBMSs Sharding

● Scaling out (horizontal scaling) based on data partitioning, i.e. dividing the database across many (inexpensive) machines

Trang 12

Scaling RDBMSs is Expensive and Inefficient

[http://www.couchbase.com/sites/default/files/uploads/all/whitepapers/NoSQLWhitepaper.pdf]

Trang 13

Not Only SQL

Trang 14

Trang 15

NoSQL History

● It was first used in 1998 by Carlo Strozzi to name his relational

database that did not expose the standard SQL interface

● The term was picked up again in 2009 when a Last.fm develper, Johan Oskarsson, wanted to organize an event to discuss opensource

distributed databases.

● The name attempted to label the emergence of a growing number of

nonrelational, distributed data stores that often did not attempt to

provide ACID

Trang 16

Trang 17

NoSQL Cost

Trang 18

SQL vs. NoSQL

Trang 19

Consistency

 Single storage image. Informally, after an update completes, any subsequent

access will return the updated value.

Trang 20

Trang 21

Quorum Model

● N : the number of nodes to which a data item is replicated.

● R : the number of nodes a value has to be read from to be accepted.

● W : the number of nodes a new value has to be written to before the write operation is finished.

● To enforce strong consistency: R + W > N

Trang 22

Quorum Model

● N : the number of nodes to which a data item is replicated.

● R : the number of nodes a value has to be read from to be accepted.

● W : the number of nodes a new value has to be written to before the write operation is finished.

● To enforce strong consistency: R + W > N

R = 3, W = 3, N = 5 R = 4, W = 2, N = 5

Trang 23

Relaxing ACID Properties

● The largescale applications have to be reliable : availability + redundancy

● These properties are difficult to achieve with ACID properties.

● The BASE approach forfeits the ACID properties of consistency and isolation in favour

of availability, graceful degradation, and performance

Trang 24

Trang 25

CAP Theorem

● C onsistency : how a a system is in a consistent state after the execution of an operation.

Trang 26

CAP Theorem

● C onsistency : how a a system is in a consistent state after the execution of an operation.

Trang 27

Trang 28

Dynamo

Trang 29

Trang 30

Trang 31

Trang 32

Trang 33

Trang 34

Trang 35

Load Imbalance

● Consistent hashing may lead to imbalance

Trang 36

Load Imbalance

 Node identifiers may not be balanced

Trang 37

Load Imbalance

 Node identifiers may not be balanced

 Data identifiers may not be balanced

Trang 38

fatemeh.mp3

jim.mp3

Trang 39

Trang 40

Trang 41

Trang 42

Replication

● To achieve high availability and durability, Dynamo replicates its data

on multiple hosts

● The list of nodes that is responsible for storing a particular key is called the preference list

5

Jim

Cosmin

Tallat Seif

Fatemeh

Trang 43

5

Jim

Cosmin

Tallat Seif

Fatemeh

Tallat

Seif

Trang 44

Trang 45

Trang 46

Data Versioning

● Version branching can happen due to node failures, network

failures/partitions, etc

 Target applications are aware that multiple versions can exist.

 If causal : older version can be forgotten

 If concurrent : conflict exists, requiring reconciliation

● A put requires a context, i.e., which version to update

Trang 47

Trang 48

Trang 49

Trang 50

Trang 51

Trang 52

Trang 53

Trang 54

Adding Node

● A new node X added to system

 X is assigned key ranges w.r.t. its virtual servers

 For each key range, it transfers the data items

Trang 55

Trang 56

Trang 57

Trang 58

Trang 59

Trang 60

BigTable

Trang 61

Trang 62

Trang 63

Table Model

● Distributed multidimensional sparse map

Trang 64

Table Model Rows

● Every read or write in a row is atomic

● Rows sorted in lexicographical order

“com.cnn.www”

Trang 65

Trang 66

Trang 67

Tablets: Pieces of a Table

● A table starts as one tablet

 As it grows, it it split into multiple tablet.

Trang 68

Trang 69

Trang 70

Trang 71

Trang 72

Trang 73

Trang 74

Trang 75

Major Components

● Tablet server

● Master server

● Client library

Trang 76

Trang 77

Trang 78

Major Components – Client Library

● Library that is linked into every client

● Client data does not move though the master

● Clients communicate directly with tablet servers for reads/writes

Trang 79

Highlevel Structure

Trang 80

Trang 81

Trang 82

Trang 83

 Master is responsible for finding when tablet server is no longer serving

its tablets and reassigning those tablets as soon as possible.

Trang 84

Trang 85

tablet log

sstable sstable

GFS

Memeory

Write operations are logged

Recent updates kept sorted

in memory

Recent updates kept sorted

in memory Memtable and sstables are merged to

serve a read request

Trang 86

request

Tablet server

Tablet server Tablet server

Client

req ues t

response

Trang 87

Trang 88

Trang 89

Trang 90

Cassandra

Trang 91

Trang 92

Trang 93

Trang 94

Any Questions?

Định dạng
Số trang	94
Dung lượng	1,83 MB