Social Metagrid Agents

1
Social Metagrid Agents

• An Approach for Flexible Resource Allocation in
  Heterogeneous Grid Middlewares
• Gabriele Pierantoni, B.Coghlan, E. Kenny, S.
  Maad, O. Lyttleton, G. Quigley.
• Trinity College Dublin

2
Contents

• Problem statement
• Philosophy
• Architecture
• Use case

3
Problem statement
Social Metagrid Agents
Different policies
Different Actors
Different technologies
Multiple Middlewares
Resource Allocation
Multiple Scenarios
Different languages
Competition/ Cooperation
Different characteristics
4
Social Metagrid Agents

• Social As they can assemble in societies
  where cooperation and competition take place to
  allocate and share resources.
• Metagrid As they live in a Metagrid
  environment that encompasses different
  middlewares
• Agents As they expose different behaviours and
  take decisions based on their environment.

5
The production layer
Service
Service
Service
Service
Service
Service
6
The social layer
Owner
Owner
Owner
Service
Service
Service
Service
Service
Service
7
Social and Production Agents

• To allow this scenario we designed two different
  kinds of agents production agents (PMA) and
  Social Agents (SMA).

S.M.A.
These agents implement the social layer of the
architecture
P.M.A.
These agents implement the production layer of
the architecture
8
Different Social Interactions

• There are many different ways in which SMAs can
  interact to create a production chain.

S.M.A.
S.M.A.
S.M.A.
P.M.A.
P.M.A.
P.M.A.
P.M.A.
Company
Market
9
The Metagrid Architecture
User Interface
Legacy Grid
Border Service
MetaGrid Service
Legacy Grid
Border Service
MetaGrid Service
Transport Layer
Legacy Grid
Border Service
MetaGrid Service
10
The Agents architecture

• The architecture can be divided into
• Java Brains Implement the behaviour.
• Agent Skeletons Allow communication, etc..
• Grid Bodies Actuators and sensors towards Grids.

Agent Skeleton
Agent Skeleton
Java Brain
Java Brain
Grid Body
Grid Body
11
The Agents Architecture
Request
P
Response
SP
SP
SP
12
A Use case.

• This use case consists in the submission of a
  series of jobs to different grid middlewares.
• The interaction of the different jobs is
  expressed with Directed Acyclic Graphs
• The workflow engine unfolds these graphs to
  submit jobs with the right sequence.
• SMAs allocate the resources needed for the
  submission.

13
A Use Case
L1
W1
G1
W2
G2
14
Use case implementation
S.M.A.
User Interface
Job Exchange
LCG2
P.M.A.
S.M.A.
Transport Layer
GT4
P.M.A.
WORK FLOW
Security Exchange
P.M.A.