Distributed Systems Exam 1

Distributed System is a

Networked computer system in which processes & resources sufficiently spread across multiple computers to appear as a single coherent system to users.

Centralized Systems

One node controls all resources and processes

Decentralized Systems

Multiple nodes operate inependently OR still operate independently but can request resources from each other

Distributed Systems

Nodes cooperate and share resources with each other

Transitioning from decentralized to a distributed system requires:

increasing node connectivity

Transparency is when

The system hides its distribution nature from users.

Types of transparency

• Access

• Location

• Migration

• Replication

• Concurrency

• Failure

Access Transparency

Hide differences in data representation & how an object is accessed

Location Transparency

Hide where an object is located

Migration Transparency

Hide that an object may be moved to another locationR

Replication Transparency

Hide that an object is duplicated

Concurrency Transparency

Hide that an object night be shared by multiple users

Failure Transparency

Hide the failures from a user

Openess

The idea that systems should support iteroperability, portability, an extensibility.

Also, system must conform to well defined interfaces.

Interoperability

The extent in which two systems/components from different manufacturers can co-exist & work together by merely relying on each other’s services as specified by a common standard.

Portability

The extent an application developed for a distributed system can be executed without modification on a different system that implements the same interfaces

Extensibility

To easily configure the system out of parts or add new components.

3 types of Scalability

• Size Scalability

• Geographical Scalability

• Administrative Scalability

Size Scalability

Easily add users and resources to the system without notably effecting performance

Size Scalability Issues

• Computational capacity limited by CPUs

• Storage capacity including transfer rate between CPUs & disks

• Network between user & the centralized service

Geographical Scalability

Users and resources are far apart, but significant communication delays aren’t noticed

Geographical Scalability Issues

• Pre-existing systems rely on synchronous communication

• Communication in wide area networks is less reliable than LAN

• Wide area systems typically have limited facilities for multi-point communication

Administrative Scalability

An easily manageable system despite spanning multiple organisations.

Administrative Scalability Issues

Conflicting policies of resource usage, management, and security

Dependability

In order for a component to provide services to a client it may require services from other components

The 4 requirements of dependability

• Availability

• Reliability

• Safety

• Maintainability

Availability

The idea that a system is fully functioning and able to provide all expected services to its users.

Reliability

The idea that a system will run continuously for extended periods of time without failur

Safety

If a system were to temporarily fail no catastrophic event happens as a result.

Maintainability

How easily a failed system can be repaired.

Fault Tolerance has 3 metrics:

1. Mean Time to Failure

2. Mean time to Repair

3. Mean Time Between Failures

Security

Dependable Systems are also required to provide security especially in regards to confidentiality and integrity

5 Points of Security

• Confidentiality

• Integrity

• Authentication

• Authorization

• Trust

Confidentiality

Only authorized access to data

How is confidentiality enforced?

Use encryption to protect confidential data & employ secure communication protocols to ensure authorized access.

Integrity

Ensure modifications are done by authorized users

How is integrity enforced?

Use hashes & digital signatures to check for tampering.
Use controlled update mechanisms & logging to verify authorized modifications.

Authentication

Verify Identity

How is authentication enforced?

Use secure credentials such as passwords and public key infrastructures to verify users.

Authorization

Check rights to access/modify an object

How to enforce authorization?

Use access control list, role based access control to restrict access.

Trust

Confidence the other points are accurate

6 Performance Metrics

• Mean Time to Failure (MTTF)

• Mean Time to Repair (MTTR)

• Mean Time Between Failures (MTBF)

• Utilization (U)

• Throughput (X)

• Response Time (R)

Utilization (U)

The fraction of time that a service is busy.
# of Incoming requests/# of processed requests

Throughput (X)

Average number of request processed per unit of time
Number of incoming requests that have been recieved by the server

Response Time (R)

How long a service takes to process a request including time in queue
1/(1-U)

4 Security Mechanisms

• Symmetric Cryptosystem

• Asymmetric Cryptosystem

• Digital Signature

• Secure Hash Function

Symmetric Cryptosystem

Uses the same key for encryption and decryption making it fast but requires secure key sharing.

Asymmetric Cryptosystem

Uses public/private key pairs where the public key encrypts and the private key decrypts. Enables secure communication without key sharing

Secure Hash Function

Maps input data to fixed length output making it sensitive to input changes, useful for integrity checks.

3 Key Terms in Fault Handling

• Failure

• Error

• Fault

Failure

When a system cannot complete a promised service
Ex: Crashed Program

Error

Part of system state that can lead to failure

Ex: Messed up syntax

Fault

The cause of an error

Ex: Sloppy Coding

4 Fault Handling Techniques

• Fault Prevention

• Fault Tolerance

• Fault Removal

• Fault Forecasting

Fault Prevention

Prevent the occurence of a fault

Fault Tolerance

Build a component to hide the occurrence of a fault from users.

Fault Removal

Reduce the presence, number, or security of a fault

Fault Forecasting

Estimate current presence, future incidents, and consequences of faults

6 Types of System Classifications

• Cluster Computing

• Grid Computing

• Pervasive System

• Mobile Computing

• Sensor Networks

• Edge Computing

Cluster Computing

Homogenous systems on LAN

• Same OS & near identical hardware

• Single or tightly couple managing nodes

Grid Computing

Heterogenous systems across WANs

• Can have different OS & Hardware

• Dispersed across multiple organizations

• Can easily span a WAN

Pervasive System

Embedded mobile computing, context aware

• Nodes are small and mobile

• Often nodes are embedded in a larger system

• System naturally blents into the user’s environment

Three subtypes of pervasive systems

• Ubiquitous

• Mobile

• Sensor

Mobile Computing

Devices as clients of cloud services

• Device location is expected to change

• Maintaining continuous communication can lead to problems

Sensor Networks 

Many small low power nodes

• Many nodes (10s-1000s)

• Simple nodes (small memory/compute/communication capabilities)

• Often battery powered or powerless

Edge Computing

Processing new data sources

• Uses mobile computing as a base but instead of directly connecting ot the cloud it connects to a mor elocal data center.

False Assumptions

• The network is reliable

• There is one adminstrator

• Transport cost is zero

• Topology does not change

• Latency is zero

• Bandwidth is infinite

Architectural Styles

Define how components interact via connectors and data exchange

Connector 

Mediates communication, coordination, and cooperation between components

Layered Architectures

Organize systems into layers with distinct responsibilities

OIS Model

The first layered model that used layers, but some aspects were vague

Database Architectuers

Use 3 layers:

• Interface Layer

• Processing Layer

• Data Layer

Interface Layer

User/external application interactions

Processing Layer

Functions of an applicationDa

Data Layer

Persistent Storage of data

Object Based Style

• Components = Objects

• Communication via method calls

• Encapsulation of Data and behavior

  • Data can be modified without revealing internal workings

RESTful Architecture is optipized around

Resources

RESTful Architecture

1. Resources are identified through a single naming scheme

2. All services offer the same interface

3. Messages sent to or from are fully self-described

4. After execution the component used forgets everything about the caller

Coupling Dimensions

• Referentially Coupled

• Temporally Coupled

Referentially Coupled

Explicit referencing in communication.
Example: Email address

Temporally Coupled

Whether or not systems are insync

Example: Instant messaging back and forth

Linda Tuple Space

A shared data space that is temporally and referentially decoupledT

Middleware

Holds commonly used components and function that don’t need to be implemented separately.
The “OS” of distributed systems.

Middleware Functions

Communication, Naming, Security, Persistance

Legacy Integration

In short use wrappers/adapters 

1 on 1 wrapper time complexity

O(N)²

Broker wrapper time complexity

O(N)

Client Server Model:

Servers offer services

Clients use services

Clients and servers can be on different machines

Clients follow requeest reply model regarding using services

Symmetric Architectures

• Peer 2 Peer

• Structures P2P

• Unstructured P2P

• Super Peer Networks

• BitTorrent

Peer 2 Peer

All nodes are equal

Structured P2P

Use hash based indexing

Unstructured P2P

Random graphs, flooding, random walksSu

Super peer networks 

Introduce hierarchy for efficiency

BitTorrent

Tracker Based swarm coordination chunk trading among peers

Hybrid Architectures

• Cloud computing layers

• Edge Computing

• Blockchain

Cloud Computing layers

Hardware → Infrastructure → Platform → Application