Transparency in a Distributed System

 ch1

Access - Hide differences in data representation and how a resource is accessed

Location - Hide where a resource is located

Migration - Hide that a resource may move to another location

Relocation - Hide that a resource is mored to another location while in use

Replication - Hide that a resource is replicated

Concurrency - Hie that a resource may be shared by several competitive users

Failure - Hide the failure and recovery of a resource

Scalability Limitations

ch1

1) Centralized Services

Services are implemented by means of only a single server running on a specific machine in a distributed system. BOTTLENECK

2) Centralized Data

Data located on a single server would also create the same kind of bottleneck. But how can we get around the need to keep 50 million addresses in one location without duplicate copies/syncing issues?

3) Centralized Algorithms

An algorithm that collects data from all sites, sends to a single machine for processing, then distributes results should be avoided.

 Characteristics of Decentralized Algorithms

(Part of Scalability Problems)

1) No machine has comlete information about the system state

2) Machines make decisions based only on local information

3) Failure of one machine does not ruin the algorithm

4) There is no implicit assumption that a global clock exists

Scaling Techniques

Persistent Data

At the Data Level...

Data is persistent.....

even if no application is running, data will be stored somewhere for next use.

chap 2

Pitfalls When

Developing Distributed Systems

 * •The network is reliable.

Scalability Limitations

Centralized Services - A server for all users

Centralized Data - A single telephone book

Centralized Algorithms - Doing routing based on complete information

Application Layering

The User Interface Level

Contains all that is necessary to directly interface with the user, such as display mangement

The Processing Level

Typically contains the applications.

The Data Level

Manages the actual data that is being acted on.

IDEMPOTENT

When an operation can be repeated multiple times without harm.

For example, if you ask "how much money do I have?" this is idempotent.  If you withdrawal $20, then it is not.

Architectural Styles

Layered Architectures

Object-Oriented Architectures

Data-Centered Architectures

Event-based Architectures

ch2

A system that monitors it own behavior and takes appropriate measures to correct itself when needed.

ch2

Software Architectures

Tell how the various software componenets are to be organized and how they should react within the organization of a distributed system.

ch2

System Architectures

The instantiation of software components on real machines.

ch2

 Vertical Distribution

ch 2

more goes here.........

The characteristic feature of vertical distribution is that it is achieved by placing logically different components on different machines.

Distributed Hash Table (DHT)

The most-used deterministic procedure in a structured peer-to-peer architecture is to organize the processes through a DHT.

MORE HERE......

ch2

Peer-to-Peer Systems

A class of modern system architectures that support horizontal distribution are known as peer-to-peer systems.  At a high level, the processes that constitute this type of system are equal.  ie: Every function that needs to be carried out is represented by every process that constitues the distributed system.

?????                ch2

Virtualization

ch3

Virtualization allows an application, and possibly also its complete environment including the operating system, to run concurrently with other applications,

but

highly independent of the underlying hardware and platforms,

leading to a high degree of portability.

stub

(in distributed computing)

ch4

A piece of code used for converting parameters passed during a RPC.

The main idea of an RPC is to allow a local computer (client) to remotely call procedures on a remote computer (server). The client and server use different address spaces, so conversion of parameters used in a function call have to be performed, otherwise the values of those parameters could not be used, because of pointers to the computer's memory pointing to different data on each machine. The client and server may also use different data representations even for simple parameters (e.g., big-endian versus little-endian for integers.) Stubs are used to perform the conversion of the parameters, so a Remote Function Call looks like a local function call for the remote computer.

Control Abstraction

ch4

Layered Architecture

ch4

Layered architecture partitions the concerns of thee application into stacked groups (layers).

It is a technique used in designing computer software, hardware and communications in which system or network components are isolated in layers so that changes can be made in 1 layer w/o affecting the others.

Level of Abstraction

ch4

Abstraction: a concept or idea not associated with any specific instance.

Control Abstraction: Abstraction of actions

Data Abstraction: Abstraction of data structures

The mechanism and practice of ABSTRACTION reduces and factors out details so that one can focus on a few concepts at a time.

NAMES

ch 5

Refer to entities

3 types of NAMES:

ADDRESS

IDENTIFIER

HUMAN-FRIENDLY

Level of Abstraction

misc

Abstraction: a concept or idea not associated with any specific instance.

Control Abstraction: Abstraction of actions

Data Abstraction: Abstraction of data structures

The mechanism and practice of ABSTRACTION reduces and factors out details so that one can focus on a few concepts at a time.

Scalability Problems

ch 1

.....of decentralized algorithms

* No machine has complete information about the system state

* Machines make decisions based only on local information

* Failure of one machine does not ruin the algorithm

* There is no implicit assumption that a global clock exists

THINK

ch1

WHERE DO YOU WANT THE DATA?

IS IT SECURE?

CAN IT BE ON THE CLIENT?

(ie: ss# should NOT be on client)

Scalability Limitations

ch 1

# 2

Centralized Services

A server for all users

Centralized Data

A single telephone book

Centralized Algorithms

Doing routing based on complete information

A Distributed System is......

ch 1

A collection of independent

computers that appears to its

users as a single

coherent system

Mutual Exclusion

ch6

* Needed when multiple processes need to smultaneously access the same resource(s).

Distributed mutual exclusion algorithms are classified into two different categories:

1) token-based solutions

2) permission-based solutions

N I C

ch??

NETWORK INTERFACE CARD

..is a hardware device that handles an interface to a computer network and allows a network-capable device to access that network.

* The NIC has a ROM chip that contains a unique #; the media access control (MAC) burned into it.

* Exists on the Data Link Layer of the OSI Model

Logical Clocks

ch6

.....what usually matters is not that ll processes agree

on exactly what time it is, but rather that the agree on

the order in which events occur.

A Note on

Ordered

Message

Delivery

ch6

1 and 2 of 2

Controversy about whether totally-ordered and causually-ordered multicasting should be provided a the middleware level vs the application handling it.

1) Middleware cannot tell what a message actually contains so only potential causality is captured. Example: 2 messages from the same sender that are completely independent will always be marked as causally rellated by middleware. This is overly restrictive and could lead to efficiency problems.

2) Not all causality may be captured. 

Example: Alice posts an article. Alice calls Bob and verbally tells him about it.  Bob posts an article without seeing Alice's post.  There is causality between Bob's posting and Alice's posting due to external communication.  This causality is not captured by the bulletin board system.

N T P

ch6

NETWORK
TIME
PROTOCOL

A timestamp is addded to the message before it is passed on the Network Interface.

        Reference

             Broadcast

                   Synchronization

(RBS)

...is a clock synchronization protocol that is quite different from the other proposals.

RBS lets the receiver synchronize, keeping the sender out of the loop.

In RBS, a sender broadcasts a reference message that will allow its receivers to adjust their clocks.

CRITICAL PATH IS MINIMIZED WITH RBS.

Pitfalls When Developing

Distributed

Systems

ch1

False assumptions.....

• The network is reliable
• The network is secure
• The network is homogeneous (transparency dictates that it is not.)
• Topology does not change (assume that it doesn't)
• Latency is zero
• Bandwidth is infinite
• Transport cost is zero
• There is one administrator

IDENTIFIER

ch5

...... a type of name

3 Properties: 

• An identifier refers to at most one entity
• Each entity is referred to by at most one identifier
• An identifier always refers to the same entity

R P C

(Remote Procedure Call)

ch 4

OSI LAYERS

vs

TCP/IP

protocol slides

What is a Layer?

ps

A layer is a collection of

conceptually similar functions

that provide services to the

layer above it and receives

services from the layer below it.

7 Layers of the

OSI Model

ps

All

People

Seem

To

Need

Data (link)

Processing

The OSI Model...

ps

The OSI Model is a way of subdividing a communications system into smaller parts called Layers.

7 - Application

6 - Presentation

5 - Session

4 - Transport

3 - Network

2 - Data Link

1 - Physical

IP Address

An IP address is a logical address. It can be changed.

(vs a MAC address that is a physical address of the hardware.  It does not change.)

Overhead

ps

Overhead associated with layers is important.

Overhead bits are nondata bits necessary for transmission

(usually as part of headers, checksums and such). 

Transparency

ch1

Access - How

Location - Where

Migration - May move

Relocation - May move while in use

Replication - has extra copies

Concurrency - May be shared by competition

Failure - Hide failure and recovery

ALMRRCF

R T P

ch 4

REAL TIME PROTOCOL

Allows timeliness over completeness for streaming video.