Manav Goyal

Introduction - Distributed Systems

Definitions
- Collection of autonomous computers linked by a Communication network and equipped with distributed system software
- Enables computers to coordinate their activities
- Example => Internet, Intranet
Characteristics
- Resource Sharing
  - Capability to use resources like hardware, software and data anywhere in the system
- Heterogeneity
  - Working with different types of OS
- Openness
  - Concerned with extensions and improvements
- Concurrency
  - Separate activities between users, independence of resources and location of server
- Scalability
  - Increase the scale of system
- Fault Tolerance
  - Cares about reliability of the system
- Transparency
  - Hides the complexity of the distributed system to the users and application programs
Application
- Finance & Commerce => E-commerce
- Information Society => Search engine, Social networking
- Creative industries and Entertainment => Online gaming, Music
- Healthcare => Patient records
- Education
- Transport and Logistics => GPS
- Science
- Environment Management => Sensors
Architecture
- Shared Memory (Tightly Coupled System)
  - Easier to program
- Distributed Memory (Loosely Coupled System)
Advantages
- Better price performance ratio
- Shorter response time and higher throughput
- High reliability against component failures
- Extensibility & incremental growth
Disadvantages
- Relevant software does not exist currently
- Security problem due to easy access to data
- Networking saturation may cause hurdle in data transfer
Issues in Designing DS
- Transparency
- Flexibility
- Reliability
- Performance
- Scalability
- Security
Types of DS
- Distributed Computing System
  - Uses group of computers that shares a common computational problem among them so as to generate an efficient results in short time span
  - Types
    - Cluster Computing
      - Group of computers are linked together in a network to perform a single task and act like a single entity
      - High availability
      - Load balancing
      - Only supports homogeneous infrastructure
    - Grid Computing
      - Every computer has a common goal and working on only one task at a time
      - Works with heterogeneous infrastructure
- Distributed Information System
  - Used to keep track of information about application running over the network which are distributed among various entities
  - Used for management and integration of business info
  - Example => Transaction processing system, Enterprise app integration
- Distributed Pervasive System
  - Used with mobile and embedded devices
  - Characterized by small battery powered device having only wireless connection, it is related to system that has lack of human control
  - Example => Home System
Open System Interconnection (OSI) Reference Model
- Trick => A Pokemon Sucked The Nectar Directly & Physically
- Physical Layer
  - Activate, Maintain and Deactivates physical connections
  - Define voltage and data rates needed for transmission
  - Digital data bits => Electrical signal
  - Decide transmission
    - Simplex
    - Half Duplex
    - Full Duplex
  - Does not perform Detection or correction of errors
- DataLink Layer
  - Synchronization and error control for the info which is to be transmitted over physical layer
  - Enable error detection
  - Adds error detection bits to the data which is to be transmitted
  - Encoded data is passed to the physical layer
  - Error detection bits are used by DDL on other side to detect & correct errors
- Network Layer
  - Mainly deals with routing
  - Routes the signals through various channels to the other end
  - Divides outgoing messages into Packets and vice versa
- Transport Layer
  - Guarantees transmission of data from one end to the other
  - Does multiplexing, splitting or segmentation of data
  - Decides if data transmission placed on Parallel or Single path
  - Data groups into smaller units
- Session Layer
  - Manages and synchronizes conversation between two different applications
  - User will establish system to system connections
  - Controls authentication, bulling and session management
  - Avoids dataloss
- Presentation Layer
  - Converts/Translates forms/syntax between sender & receiver and vice versa
- Application Layer
  - Provides manipulation of info in various ways
  - Retransferring the files on info
  - Distributing the result
  - Login/Password checking
Code Migration
Consistency and Replication
- Types of Replica
  - Permanent Replica
  - Server-initiated Replica
  - Client-initiated Replica
- Models
  - Master-Slave
  - Client-Server
  - Peer-to-Peer
Fault Tolerance
- Failure in DS means when a service cannot be fully provided
  - System faliure may be partial
- Failure Escalation: Single failure may affect other parts of a system
- Types of Failure
  - Omission Failure: Server fails to respond to incoming messages
    - Receive Omission: Server fails to receive incoming messages
    - Send Omission: Server fails to send messages
  - Timing Failure: Server response lies outside the specified time interval
  - Arbitrary Failure: Server produces arbitrary responses at arbitrary times
Distributed File System
- Files are available on several computers, which can be shared
- Supports
  - Remote Information Sharing
  - User Mobility
  - Diskless Workstations
  - Availability
- Services Provided
  - Storage
  - Truefile: Supports operation like Creating, Deleting, Modifying, ... a file
  - Name
- Features
  - Transparency: Structural, Access, Naming, Replication
  - User Mobility
  - Performance
  - Simplicity
  - Scalability
  - High Availability & Reliability
  - Data Integrity
  - Security
  - Heterogeneity