Understanding Code Mobility

Sachin Chouksey H. Qiu

Xiong

Distributed Applications Problems with traditional approaches

Technologies,architectures,methodologies.

Large scale distributed settings. Fail to provide

Configurability Scalability Customizability

Code Mobility An emerging research field. However

Lack of standard terms and concepts. Difficult to assess and compare existing

approaches. Paper presents conceptual framework. Classification

Technologies Design Paradigms Applications.

Development Environment Computer Networks

Increasing size and performance Pervasive Ubiquitous

Mobile computing World Wide Web

Problems Size of Networks (Scalability) Wireless Connectivity

Dynamic topologies Large scale diffusion into society

Customizability Flexibility Extensibility

Current proposed Approaches

Similar Approaches Remote Job submission

Process Migration

Object Migration

BUT THESE DO NOT SCALE

How is code mobility different ? Internet-scale Location Aware Programmer controlled mobility Mobility for

Load Balancing Service Customization Dynamic extension and application functionality Autonomy Fault tolerance Disconnected Operation

3D Classification Technologies

Languages & systems that provide mechanisms Used in implementation stage by developer

Design Paradigms Architectural styles used C2, P2P, Client-Server

Application Domains Classes of applications E-Commerce, Distributed information retrieval

Mobile Code Technologies Traditional Distributed System Layers

Core Operating System Network Operating System True Distributed System

For Code Mobility Computational Environments (CE) CE retains identity of the host Dynamically relocate components on different

hosts CE hosts Execution units (EUs) and resources

Mobility Mechanisms EU = Code Segment + Execution

State + Data Segment

EU moves from CE to CE

Orthogonal mechanisms Code and execution state mobility Data space management

Code and Execution State Mobility

Strong mobility Migration

Proactive / Reactive Remote cloning

Weak mobility Fetch / Ship Stand alone code / code fragment Synchronous / Asynchronous

Data Space Management Rearrange EU-Resource bindings after

migration Resource = <I,V,T> Bindings

By Identifier By Value By Type

Bindings constrain the data space management mechanisms Resource Relocation Binding reconfiguration

Current Mobile Code Technologies Agent Tcl Ara Facile Java Java Aglets M0 Mole Obliq

Safe-Tcl Sumatra TACOMA Telescript

Design Paradigms Design paradigms, or architectural

styles, define architectural abstractions and reference structures that may be instantiated into actual software architectures

Why Design Paradigms? The goal of design is the creation of a

software architecture, which can be defined as the decomposition of a software system in terms of software components and interactions among them

Software architectures with similar characteristics can be represented by design paradigms

Describe Design Paradigms Design paradigms are described in

terms of interaction patterns that define the relocation of and coordination among the components needed to perform a service

Components Components are the constituents

of a software architecture. They can be further divided into: Code components Resource components Computational components

Interactions Interactions are events that

involve two or more components, e.g., a message exchanged among two computational components

Sites Sites host components and support

the execution of computational components

A site represents the intuitive notion of location

Design Paradigms Client-Server (CS) Remote Evaluation (REV) Code on Demand (COD) Mobile agent (MA)

Comparison of the Paradigms

ParadigmBefore After

SA SB SA SB

Client-Server

Aknow-how resource

BA

know-howresource

B

Remote Evaluation

know-howResource

BA

know-howresource

B

Code on Demand

resourceknow-how

B

know-howresource

AB

MobileAgent

know-how resource -know-howresource

B

Outline Add the 1st person’s Design Paradigms Add the 3d person’s

Design Paradigms Design paradigms, or architectural

styles, define architectural abstractions and reference structures that may be instantiated into actual software architectures

Why Design Paradigms? The goal of design is the creation of a

software architecture, which can be defined as the decomposition of a software system in terms of software components and interactions among them

Software architectures with similar characteristics can be represented by design paradigms

Describe Design Paradigms Design paradigms are described in

terms of interaction patterns that define the relocation of and coordination among the components needed to perform a service

Components Components are the constituents

of a software architecture. They can be further divided into: Code components Resource components Computational components

Interactions Interactions are events that

involve two or more components, e.g., a message exchanged among two computational components

Sites Sites host components and support

the execution of computational components

A site represents the intuitive notion of location

Design Paradigms Client-Server (CS) Remote Evaluation (REV) Code on Demand (COD) Mobile agent (MA)

Comparison of the Paradigms

ParadigmBefore After

SA SB SA SB

Client-Server

Aknow-how resource

BA

know-howresource

B

Remote Evaluation

know-howResource

BA

know-howresource

B

Code on Demand

resourceknow-how

B

know-howresource

AB

MobileAgent

know-how resource -know-howresource

B

Mobile Code Applications(1) They are still immature and lack of

suitable methodologies. The benefits of mobile code are

appealing.(Mobile agent systems, Internet agents)

It is important to distinguish the implementation process and the paradigm.

Mobile Code Applications(2)Benefits:

1. Service Customization- The interfaces or services are are not statically defined.2. Supporting more flexible software deployment and maintenance.3. Automatic software upgrade without human interaction on the client side.

Mobile Code Applications(3)

Benefits:4. Autonomy of application components.

- heterogeneous communication infrastructure complexes the network and requires autonomic design.Two solutions:

Put more operations on the server

Or put more operations on the clients

Mobile Code Applications(4)Benefits:

5. The fault tolerance is much improved in Mobile code than the conventional client-server system.

6. A new protocol is needed to enhance the code migration among distributed systems.

7. Date management flexibility and protocol encapsulation.

Mobile Code Applications(5)Application Domains for Mobile Code 1. Distributed Information Retrieval2. Active Documents3. Advanced Telecommunication Services4. Remote Device Control and Configuration 5. Workflow Management and Cooperation6. Active Networks7. Electronic Commerce

A Case Study In Network Management (1) Determine if mobile code is suitable

for the specific application first. Identify the corresponding paradigms. Analyze the tradeoff for each

application functionality. Select the technology to implement

the application according to the tradeoff.

A Case Study In Network Management (2)

The network congestion traffic problem: 1. The problem to solve: Decentralizing

Network Traffic-Both SNMP and CMIP are inefficient while the network heavy congestion.- Some modifications have been made to solve the problem, such as the SNMPv2 with the proxy agents and the Remote Monitoring.

A Case Study In Network Management (3)2. Identifying the Design Paradigms

-REV paradigm can be used, which will pack up all the SNMP operations and send them on the device for the local interaction.-Only the useful target value will be sent back after execution.-The capability to retain the state across several hops will also make the whole system more autonomic.

A Case Study In Network Management (4) 3. Evaluating the Design Tradeoffs(details will not be discussed)-The mobile code is only convenient

when a set of SNMP instructions can be “packed” efficiently into mobile code.

-The choice of the paradigm is constrained by the actual values for the parameters of the application.

A Case Study In Network Management (5)

Choosing the Implementation Technology

-In principle, design paradigms and implementation are orthogonal. Sometimes, they are different.

-We also have to think about the global application development when we implement a given functionality inside it.

Summary and Conclusions Mobile code is a promising solution for

the design and implementation of large scale distributed applications

Little attention has been paid to the formulation of a sound conceptual framework for code mobility

Three classes of concepts in a proposed conceptual framework: applications, design paradigms and technologies

Strengths and Weaknesses Systematically presenting

knowledge about code mobility: motivations, concepts, principles, approaches, technologies, etc., therefore, providing a complete image

Expounding is vivid with metaphor

Relevance to Embedded Systems Many embedded systems are

resource-constrained Requirements to them are higher and

higher Code mobility technology providing a

possible solution Many embedded systems are real-

time systems that can not be stopped to do operations such as update