Towards a Universal Client for Grid Monitoring Systems Towards a Universal Client for Grid Monitoring Systems Design and Implementation of the Ovid Browser.

Towards a Universal Client for Grid Towards a Universal Client for Grid Monitoring SystemsMonitoring Systems

Design and Implementation of the Ovid Browser

M. D. Dikaiakos A. Artemiou G. TsouloupasUniversity of Cyprus

Presenter: Marios D. DikaiakosHIPS 2006

The Grid Middleware infrastructure that enables flexible,

secure, coordinated resource sharing among dynamic collections of individuals and institutions (Foster, Kesselman, Tuecke).

Enables communities (“Virtual Organizations”) to share geographically distributed resources as they pursue common goals.

Emerging Grid infrastructures characterized by: Large scale (size & geography) High complexity Heterogeneity (resources and services) Lack of central control

The Grid information problem How are individuals and organizations going to

harness the capabilities of a fully deployed Grid: Large and expanding base of resources. Huge corpus of available programs, services, and data.

Users need tools to discover and represent information about the structure configuration state of Grid resources.

Grid Information and Monitoring Services

Collect and provide information that is essential to the operation of a Grid infrastructure: Static representations of Grid-resource

characteristics; Descriptions of existing services, software,

applicable policies, and user accounts; Dynamic representations of resource status,

performance, and availability. Easy Easy andand seamless access seamless access to such informationto such information is is

necessary to lower the barriers of entry to the necessary to lower the barriers of entry to the GridGrid.

Grid Information and Monitoring Services

A variety of client systems that: support different types of information, operate on top of different underlying middleware, speak different protocols for retrieving and/or

publishing information.

Most monitoring systems publish their information on the Web; however: they do not support the view of a coherent

information space; their information is essentially represented in

tabular formats and listings.

Globus MDS

QuickTime™ and aTIFF (LZW) decompressor

are needed to see this picture.

GridICE

GridICE

MapCenter

JMX-based Infrastructure Monitoring System (JIMS)

Motivation

The discovery and retrieval of information about the status and configuration of Grid infrastructures remains a daunting experience and a major obstacle to the Grid’s wider adoption.

We need monitoring-clients that: can retrieve information from different sources,

using different protocols on the back-end, maintain the view of a coherent information space

on the end-user side.

QuickTime™ and aTIFF (Uncompressed) decompressor

are needed to see this picture.

Outline

Introduction and Motivation

Ovid: Key Concepts and Functionality

Ovid Design and Implementation

Conclusions and Future Work

Ovid overview

Supports end-user navigation inside a virtual information hyperspace, whose structure is defined by a model of the Grid: Hyperspace nodes entities of the Grid

model. Node content: retrieved dynamically from Grid

information sources. Hyper-links: represent hierarchical

containment or reference relationships between interlinked entities of the Grid model

QuickTime™ and aTIFF (LZW) decompressor

are needed to see this picture.

Ovid: key aspects

Navigational primitives designed to cope with network disorientation and information overloading;

A small set of core graphical Ovid views, i.e. visual abstractions of Grid information;

Support for embedding and implementing hyperlinks connecting related entities represented within different information views;

A plug-in mechanism, for the seamless integration with Ovid of third-party monitoring clients;

A modular software design (model-view-controller architecture), for the easy integration of different visualization algorithms.

Ovid Views

Spatial hypertext maps: Store attributes of Grid entities Contain statically embedded hyperlinks Support the dynamic installation and invocation

of external hyperlinks, retrieving content from third-party monitoring services

Supported views: VO-Sites Network Topology

VO-Sites View (CrossGrid)

Computing Element

StorageElement

Worker Node

Navigation bar

VO-Sites View (EGEE-dteam)

QuickTime™ and aTIFF (Uncompressed) decompressor

are needed to see this picture.

Network-topology View

Querying a Grid Site

History Manager

Bookmarks Manager

Ovid’s Search Interface

Ovid: Plug-in Mechanism

Mapcenter’s Plug-in

Plug-in Manager

Outline

Introduction and Motivation

Ovid: Key Concepts and Functionality

Ovid Design and Implementation

Conclusions and Future Work

Model-View-Controller Design Paradigm

Divides functionality of OO applications into: Model: contains the data sources in which all data

manipulation and processing operations take place. View: contains all the “views” derived by the corresponding

model; a view is a visual and/or textual representation of data

Controller: handles user interaction, interprets user requests into messages sent to the Model

We merge View and Controller into one category, the Delegate: a design pattern introduced by Sun in its Swing components.

In Ovid, each information source is managed by a different Model-Delegate module.

Design Diagram

QuickTime™ and aTIFF (LZW) decompressor

are needed to see this picture.

Internal StateInternal StateFront-end modulesFront-end modulesBack-end modulesBack-end modules

User Context The virtual “location” of a user during his

navigation is represented by a Context object: Active Virtual Organization Selected Grid resource Type of selected resource Active Model-Delegate entity User security certificate

Context object is used for the proper interpretation of end-user interactions with Ovid.

Context changes are registered by the History Engine.

Navigation Support

Navigation operations (next, back, refresh, search) managed by the Navigation Manager in collaboration with the History Engine or the Plug-in Manager.

Hyperlinks: clickable object-geometries embedded in Ovid views and associated with some hyperlink resolver.

Ovid Plug-ins Introduced to support the retrieval and display of

information derived from a variety of Grid information sources (monitoring systems). No “hard-wiring” of the code that handles the

protocols and information encoding of specific monitoring systems.

Ovid plug-ins are small, pluggable components written in JAVA that implement the ConnectionClass  interface provided by Ovid.

The Plug-in Manager of Ovid supports the easy download, installation and configuration of Ovid plug-ins developed by third-parties.

Implementation Details

Implemented in JAVA.

Can be installed as a standalone jar file.

Includes a software cache that allows the system to run and present data even in the absence of network connectivity.

Tested with success both on CrossGrid and EGEE test-beds.

Conclusions End-user navigation inside large information

spaces that represent the configuration, the capabilities and the state of open, large-scale computational infrastructures is important and challenging.

Ovid represents an approach in tackling this challenge in the context of Grid infrastructures.

Ovid supports: Navigation through spatial hypertext maps that

represent graphically a model of Grid infrastructures.

The easy integration of external information sources through the plug-in mechanism.

Future Steps Support the definition and submission of Grid jobs

using a drag-and-drop graphical interface. Improve the visualization algorithms (using

GraphViz libraries). Investigate the use of ontology languages (OWL)

for the internal representation of the Grid model. Extend the “search” functionality by integrating

external searching facilities (e.g. for software components).

Provide more plug-ins. g-Eclipse.