Transcript Remote execution

Distributed Process Implementation
Hima Mandava
OUTLINE

Logical Model Of Local And Remote
Processes
 Application scenarios
Remote Service
Remote Execution
Process Migration
 Mobile agent
 Process migration vs. mobile agents
Logical model of local and remote processes



Client server model similar to RPC
implementation.
It assumes that there are front-end stub processes
to facilitate the creation of and communication
between processes on different machines.
All subsequent communication between local and
remote processes is supported indirectly through
stub processes.
Logical model of local and remote processes

CLIENT
SERVER
Local Process
Remote Process
Stub Process
Stub Process
Stub Processes at the server site interprets the messages
from the client stub and invokes respective operations.
Significant Application Scenarios
Depending on how request messages are
Interpreted, there are 3 application scenarios:

Remote service: The message is interpreted as a
request for a known service at remote site.
 Remote execution: The messages contain a program
to be executed at the remote site.
 Process migration: The messages represent a
process being migrated to the remote site for
continuing execution.
Remote service
Its primary application is resource sharing in distributed
systems. A request message for service can be
generated at 3 different software levels.

As remote procedure calls at language level.
 As remote commands at the operating system level.
 As interpretive messages at application level.
Remote service



As remote procedure calls at language level.
It is service oriented and provides both access and location
transparency.
As remote commands at the operating system level.
Some frequently used commands that are built in part of
command shell like rcp in unix are used to invoke remote objects.
As interpretive messages at application level.
Messages are interpreted by user-defined daemon processes. A
good example is file transfer protocol ftp.
Primary implementation issues are I/O redirection and security in
Remote service.
Remote execution





It is spawning of a process at a remote host.
Assumes remote operation initiated by client is created
by client.
The message sent from client to a server is a client
program to be executed at the server.
It is a processor-pool model for distributed
computation.
Implementation of remote execution is more complex.
Remote execution
Additional implementation issues are:




Load sharing algorithms
Location independence
System heterogeneity
Protection and security
Load sharing algorithms

A remote host is selected first.
 For practical implementation each process server
maintains a list of registered hosts.
 Registering and deregistering services can be
implemented by broadcasting.
 Selection process can also go through a centralized
broker process.
 If the client’s resource requirements are met and is
authenticated, server grants permission for remote
execution.
 Finally client initializes the process forked at the
remote site.
Location independence




Coordination is required among processes created by
remote execution to accomplish a common task.
It is necessary to support a logical view for the
processes.
Each remote process is represented by an agent
process at the originating host.
The file system of originating process is used to
provide a common logical computer view for the
processes.
System heterogeneity




If a remote execution is to be invoked on a
heterogeneous host, it is necessary to recompile the
program.
Overhead may be too high to justify remote execution.
As a solution canonical machine-independent
intermediate language is used for program execution.
To translate data , an External Data Representation
(XDR) is integrated into remote execution facility.
Protection and Security




It is safer to accept remote execution only in source or
intermediate codes.
The language could be restricted to exclude potential
problems such as use of pointers and multiple
inheritance.
Run-time checking of parameters and stack overflow
is also necessary.
In distributed systems protection and security issues
for remote execution are topics of ongoing research.
Process migration




The load-sharing model is further extended to allow a
remote execution to be preempted and moved to
another remote host.
Since the target process for migration is preempted ,
its state information must also be transferred.
State information consists of computation state and
communication state .
The computation state is the information necessary to
save and restore a process at the remote site and
communication state is the status of process
communication links and messages in transit.
Process migration
Time
Source Site
Destination Site
Process P1
:
:
:
:
Execution
suspended
Freezing
time
Transfer of
control
Execution
Resumed
:
:
:
:
Process P1
Link redirecting
Total Freezing
Pre transferring
Source
node
Source
node
Suspended
Freezing
time
Destination
node
Destination
node
Migration
decision
Migration
decision
Transfer of
address space
Transfer of
address space
resumed
Merits:
easy implementation
Demerits: long delay time
Suspended
Freezing
time
resumed
Merits:
freezing time reduce
Demerits: total time extended
Message forwarding
Three types of messages:

1. Received when the process execution is stopped on the source
node and has not restarted on the destination node.

2. Received on the source node after the execution started on
destination node.

3. Sent to the migrant process after it started execution on
destination node.
Message forwarding
Resending messages
Origin
Receiver
Sender
Send
ResendMigrate
Dest 1
Resend again
Migrate again
Dest 2
•Messages of type 1 and 2 are either
dropped or negatively acknowledged.
•The sender is notified and it needs to
locate the migrant process
Message forwarding
Ask origin site
Origin
Sender
Send
Receiver
•Origin node keeps the info on
the current location of the
process created there.
Migrate
Forward
Dest 1
Migrate again
Dest 2
•All messages are sent to origin
which forwards them to migrant
process.
Message forwarding
Link traversal
Origin
Sender
Send
Send
Receiver
Forward
Link
Migrate
Dest 1
Forward
Link
Migrate again
Dest 2
•Messages of type 1 are queued and
sent to destination node as part of
migration procedure.
•Link is left on source node to
redirect messages of type 2 and 3.
•Link contains the system-wide
unique id of a process and its last
known location.
Message forwarding
Link Update
Origin
Sender
Send
Receiver
New location
Migrate
Send
New location
Send
Current location
Dest 1
•During the transfer, the source node
sends the notification (link update) of the
transfer to all the nodes to which the
process communicates.
•Messages of type 1 and 2 are forwarded
by the source node
Migrate again •Messages of type 3 are sent directly to
the destination node
Dest 2
Mobile agents


A mobile agent is a process that can transport its state
from one environment to another, with its data intact,
and be capable of performing appropriately in the new
environment. Mobile agents decide when and where to
move.
Common applications include:
Resource availability, discovery, monitoring
Information retrieval
Network management
Dynamic software deployment
Mobile agents





Mobile agents are agents that can physically travel across a
network, and perform tasks on machines that provide agent
hosting capability.
This allows processes to migrate from computer to computer, for
processes to split into multiple instances that execute on different
machines, and to return to their point of origin.
Unlike remote procedure calls, where a process invokes
procedures of a remote host, process migration allows executable
code to travel and interact with databases, file systems,
information services and other agents.
Mobile agents can decide when and where to move next.
When a mobile agent decides to move, it saves its own state and
transports this saved state to next host and resume execution
from the saved state.
Process Migration vs. Mobile Agents
Process migration
Mobile agents
Navigational
Autonomy
Migration decision is
made by system.
Agents decide where and
where to go
Code Execution
Programs are fully
compiled and executed
in native mode.
Most agents are coded in
Java and are interpreted
by their execution engine.
Strong/Weak
Migration
Execution is resumed
where it has been
suspended.
Java-based agents
resume their execution
from the top of a given
method.
I/O State
Long-term I/Os are
forwarded to processes
migrated to the
destination.
Agents relinquish I/O
connections every time
they depart for their next
destination.
Research work




Freeze Free algorithm for process migration to reduce
process migration latency.
A dynamic scheduling mechanism that considers
migration cost as well as other conventional influential
factors for performance optimization in a shared,
heterogeneous environment.
Extended elementary object system (EEOS), we
developed a formal model for a generic secure mobileagent system.
Redirection Algorithms for Load Sharing in Distributed
Web-server Systems.
Future work



Protection and security issues for remote execution in
distributed systems.
Improvement of Performance and scalability in mobile
agents.
A system with transparent migration facility in
distributed processing.
Resources




Distributed operating system and algorithms
Randy chow, Theodore Johnson
Fast Dynamic Process Migration
ieeexplore.ieee.org.ezproxy.gsu.edu/stamp/stamp.jsp?tp=&arnumber=508015.
(1996)
Dynamic Scheduling with Process Migration
ieeexplore.ieee.org.ezproxy.gsu.edu/stamp/stamp.jsp?tp=&arnumber=4215370
(2007)
Redirection Algorithms for Load Sharing in Distributed Web-server Systems
ieeexplore.ieee.org.ezproxy.gsu.edu/stamp/stamp.jsp?tp=&arnumber=77
6555(1999)

Formal Modeling and Analysis of a Secure Mobile-Agent System
ieeexplore.ieee.org.ezproxy.gsu.edu/stamp/stamp.jsp?tp=&arnumber=44
04055(2008)


Presentation on Parallel & Distributed Computing
courses.washington.edu/css434/slides/Migration.ppt
Mobile agent - Wikipedia, the free encyclopedia
http://en.wikipedia.org/wiki/Mobile_agent
Thank You