Collaboration in Grid Systems Activities where collaboration important

Collaboration in Grid Systems • Activities where collaboration important • Functional Structure of a Collaborative System • Collaborative Visualization • Grid Event and Message Service

Grid Collaboration Opportunities z Collaborative Portals: (Portal is a Grid Computing or Training or ? Environment) for Computing or Education z Education Grid: Distance Education is implemented as collaborative Service (between teacher and student) on top of a education/training portal z Computational Grid: Typical examples y. Collaborative Visualization y. Collaborative Consulting (help desk) y. Collaborative selection of parameters (possibly dynamically) for job submission (Collaborative computational steering)

Collaborative Grid Collab Server Pu blish Data base Collab Server Collaborating Grid with Redundant Server Network (Industry leader Web. Ex has 120 in 6 centers) for persistence and routing (optimal software multicast) Sub scribe Peer to Peer “Illusion” Among collaborating clients

What do Collaboration Systems have ? z. Audio Video Conferencing (Hear. Me at low end; Access Grid at high end) z. Text chat; white board; instant messenger z. Secure access; group e-mail; surveys, quizzes z Shared documents or objects including y. Shared display (framebuffers) and y. Shared Export (version of shared object in some standard form like HTML SVG PDF JPEG) y. Shared Event (original object)

Publish/Subscribe Collaboration z The web is full of objects – web pages sitting on web servers – and these support asynchronous collaboration y You post a web page and I later look at it in my own time z Replacing web document by a “CGI script” or servlet (web interface to program, database etc. ) gives asynchronous general multi-tier object sharing z This is Publish/Subscribe mechanism y If add some mechanism (channel or automatic email or word of mouth) to tell viewing client when new information is posted y We build on JMS (Java Message Service) as Industry standard for publish/subscribe systems z Synchronous Collaboration provides “real-time” notification and automatic update of changed objects y. JMS seems fast enough to do real-time as well asynchronously

Collaborative Computing z Shared Display for consultants (remote control for debugging). Basic help desk mode z Visualization done by shared display or shared export (Java 3 D SVG GIF) y. Shared SVG for Flash, Adobe, Power. Point, Gateway Whiteboard and ? y. Annotations (pointers) can be shared for scalable export formats as can separately zoom and still position overlays z Job Submission done by shared Java Server Pages in Gateway and shared web pages for HTML based interfaces

Collaborative Visualization Object Filter Map Transform Broker Shared export Object Filter Map Transform Event Adapter Custom Broker Output A Input W 3 C Event Adapter Input B Output Shared z Consider a computer program (object Display Output C above) and then its output and input wend their way through multiple filters (tiers) until they are finally rendered on Master User B some sort of device: CAVE through PDA. Collaborators z One can share “object” at any stage in pipeline A and C

Research Issues: Collaborative Grids z Optimal routing and possibly dynamic server arrangement based on aggregation of topic based subscriptions z Event and Message service (an event is just a time stamped message) integrating clients and servers y. XML specification of events z Compare standard JMS with proposed GMS y. GMS XML not text based properties y. GMS supports synchronous and asynchronous collaboration y. GMS has agent based routing y. HHMS or GMS Micro-edition for Hand held clients y. GMS uses meta-objects managed My. Xo. S – Web OS based on interpreted XML (RDF statements) z All Information Service issues in access of shared resources

JMS (Java Message Service) Structure Global (distributed) Event Receptor (Queue) PDA with JMS Subscribe HHMS Publish Convert Events to JMS Java. Script Java C++ …. . PDA without JMS HHMS (Hand Held Message Service) Optimized for Performance. Assume complex (XML) processing done on “personal server”

Bandwidth of Commercial JMS

Commercial JMS Latency

Comments on JMS z Message latency varies from 3 millisec for small to 300 millisecond for 1 megabyte messages (OK as dealing with human perception time) z JMS has rather limited matching criterion for topics – replace by XML based objects in GMS z Do not use on hand helds: these managed by personal server and linked by HHMS or GMSME which is very light weight z Persistence by subscribing a database system to all topics z Natural distributed model (replication or distribution of topics) for performance and fault tolerance

Proposed GMS Model for Messages z Agent based messaging with server (broker) side intelligence and archiving Subscribes to all events to get Database persistence Server Message Input Queue Agent combines Message Properties Subscriber Requests to decide on message recipient GMS: Grid Message Service can be built on commercial JMS Publishers All events in XML Central Knowledge Subscriber Profiles Set via GMS Subscribers

My. Xo. S Meta. Object Event-based Computing Paradigm z All entities are Meta. Objects defined in universal GXOS XML Syntax – Meta means GXOS doesn’t really want to manage Object, just information required to find, render and share it y. Rendering includes Palm devices as well as PC’s y. Entities are people, computers, data sources, information from e-mail, scientific visualization to digitized Bible z All actions including object changes are events – all events are GXOS objects ye-mail, paging etc. are all GXOS events uniformly routed/archived etc. z All action is instantiated or recorded in messages which are events themselves z All events are archived allowing actions to be undone and to be reviewed z There is an Meta. Object Shell My. Xo. S with basic Services (copy, create, collaborate etc. ) expressed in RDF y. Discovery etc. via Jini/Napster mechanisms