Modern database

technology
CHAPTER 4
DISTRIBUTED AND NO SQL DATABASES

SARANKUMAR K - CIT CHENNAI 1

 Distributed and no sql
databases
1. Distributed Database Architecture
2. Distributed Database Storage
3. Commit Protocols – 2PC, 3PC
4. Types of No SQL databases
5. CAP Theorem
6. Mongo DB – CRUD Operations

SARANKUMAR K - CIT CHENNAI 2

 DISTRIBUTED DATABASE
ARCHITECTURE
 Client – Server Architecture  client connect to central server 
manages distributed database system
 Peer to Peer Architecture  each site in distributed database
system  connected to all other sites
 Both responsible for
 Coordinating Transactions
 Managing Data Storage
 Providing Access Control

SARANKUMAR K - CIT CHENNAI 3

SARANKUMAR K - CIT CHENNAI 4
SARANKUMAR K - CIT CHENNAI 5
 DISTRIBUTED DATABASE
storage
 Data can be stored on different sites  by following methods
like Replication and Fragmentation
 Replication  maintain multiple copies of data  stored on
different sites  for faster retrieval  fault tolerance
 Fragmentation  partitioned into several fragments  stored in
distinct sites
 Replication & Fragmentation combined  partitioned into
several fragments  system maintains several identical
replicas of each such fragment

SARANKUMAR K - CIT CHENNAI 6

 DISTRIBUTED DATABASE
storage
 Full Replication  every site contains copy of entire database
 Advantages of Replication
 Availability  failure of site containing relation r  does not
result in unavailability of r
 Parallelism  queries on r  may be processed by several
nodes  in parallel
 Reduced data transfer  relation r is available locally at
each site  containing replica of r
 Disadvantages of Replication  Increased cost of updates 
Increased complexity of concurrency control
SARANKUMAR K - CIT CHENNAI 7
 DISTRIBUTED DATABASE
storage
 Horizontal Fragmentation  relational table or schema  is
broken down into group of one or more rows  each rows get
one fragment of schema  splitting by rows
 Vertical Fragmentation  relational table or schema  is broken
down into group of one or more columns  each columns get
one fragment of schema  splitting by columns

SARANKUMAR K - CIT CHENNAI 8

 DISTRIBUTED DATABASE
storage
 Advantages of Fragmentation
 Horizontal  allows relation to be split  tuples are located 
where they are most frequently accessed
 Allows parallel processing  on fragment of relation
 Vertical  Tuple ID attribute  efficient joining of vertical
fragments

SARANKUMAR K - CIT CHENNAI 9

 Commit protocols
 Transaction executed at multiple sites  must either committed
at all sites or aborted at all sites
 Not Accepted to have transaction  committed at one site and
aborted at another
 2PC (2 Phase Commit)  protocol widely used
 3PC (3 Phase Commit)  more complicated and more
expensive

SARANKUMAR K - CIT CHENNAI 10

 2 phase commit protocol
 One coordinator  responsible for initiating protocol
 Other entities  called participants
 If coordinator or participants  unable to commit  all parts of
transaction are aborted
 Two Phases
 Phase 1: Coordinator sends “Request to Prepare?”  to all
participants in transaction  Vote Collection Phase
 Phase 2: Implement decision “Commit Initialised” to all
participants  Decision Phase

SARANKUMAR K - CIT CHENNAI 11

SARANKUMAR K - CIT CHENNAI 12
 3 phase commit protocol
 One coordinator  responsible for initiating protocol
 Other entities  called participants
 If coordinator or participants  unable to commit  all parts of
transaction are aborted
 Three Phases
 Phase 1: Coordinator sends “Request to Prepare?”  to all
participants in transaction  Vote Collection Phase
 Phase 2: Coordinator sends “Prepare to Commit?”  to all
participants in transaction  Dissemination Phase
 Phase 3: Implement decision “Commit Initialised” to all
participants  Decision Phase
SARANKUMAR K - CIT CHENNAI 13
SARANKUMAR K - CIT CHENNAI 14
 Commit protocols
 Transaction executed at multiple sites  must either committed
at all sites or aborted at all sites
 Not Accepted to have transaction  committed at one site and
aborted at another
 2PC (2 Phase Commit)  protocol widely used
 3PC (3 Phase Commit)  more complicated and more
expensive

SARANKUMAR K - CIT CHENNAI 15

 Types of nosql Databases
 Key value pair based

 Column based

SARANKUMAR K - CIT CHENNAI 16

 Types of nosql Databases
 Document Oriented

 Graph based

SARANKUMAR K - CIT CHENNAI 17

 Cap theorem
 Consistency  consistent execution of operation
 Availability  should be always available and responsive
 Partition Tolerance  system should continue to function  even
if communication among servers is not stable
 CA – Consistency & Availability
 CP – Consistency & Partition Tolerance
 AP – Availability & Partition Tolerance

SARANKUMAR K - CIT CHENNAI 18

Cap theorem

SARANKUMAR K - CIT CHENNAI 19

Mongo db - crud operations
 Create  insertOne(), insertMany()
 Read  find(), findOne()
 Replace  updateOne(), updateMany, replaceOne()
 Delete  deleteOne(), deleteMany()

SARANKUMAR K - CIT CHENNAI 20