Evaluating Collaborative Infrastructure for CSCW Apps Grid Web Services

Evaluating Collaborative Infrastructure for CSCW Apps Grid/Web Services COMP 790, Fall 2006 Collaborative Systems

References • Co. Frame: A framework for CSCW Applications based on Grid and Web Services – Jinlei Jiang et al, Proceedings of IEEE International Conference on Web Services (ICWS 2005) • Developing Collaborative Editing Applications using Web Services – Muhammad Younas, Rahat Iqbal, Coventry University, Coventry, UK • Improving Network Efficiency in Real-Time Groupware with General Message Compression – Carl Gutwin et al CSCW 2006 • A Collaborative Infrastructure for Aerospace Design – http: //ic. arc. nasa. gov/people/filman/text/darwin/ace. pdf • A Collaborative Infrastructure for IM applications – http: //www. imlogic. com/ • A Collaborative Infrastructure for Scalable and Robust News Delivery – http: //doi. ieeecomputersociety. org/10. 1109/ICDCSW. 2002. 1030843

Collaborative Infrastructure • CSCW Applications and Infrastructure – Are they the same ? • What is a CSCW Application: – N-user software – Example: Chat Tool

CSCW Infrastructure • Higher Level of abstraction that provides application developers a framework to build CSCW applications. • Example: – We have already used one: – “Sync”: A Java based framework for developing collaborative applications

Evaluating Criteria Systems Flexibility (Automation) - Awareness Support - Directory Service - Communication Programming Model (Environment) Programming Cost (Transparency) Consistency - Merging Fault Tolerance Resource Optimization - Message Compression - Resource Replication Policy Sync Co. Frame WS

Features Supported / Flexibility • News. Wire collaborative content delivery system – Deals with real-time delivery of news items – Reduces Compute/Network load • DARWIN – Collab Infrastructure for Aero. Space Design • Issues such as: wind tunnel testing data – Focus: • Storing/Indexing data • Manage visualizations for stored/derived data • IMLogic – IM Application development infrastructure • Drawbacks – Specific to one application. – Deal with some core issues but not all • Alternatives: – Sync (Framework for Collaborative Applications) – Web Services and Grid Computing

Features supported – Grid/Web • Provide Core Services and additional services – Core Services: • Communication – Cooperative Message Bus (CMB) – RMI (Sync Equivalent) – SOAP (Simple Object Access Protocol) – Additional Services: • Process Management Service • Awareness Support Service • Interaction-Aided Service – Directory Service » UDDI (Universal, Description, Discovery and Integration) » Central Registry (similar to UDDI) » RMI registry (Sync) – Persistence Visual Cooperation space facility • Easier to add new functionality as it gets added as another service in the framework

Programming Model (Environment) • Grid Computing – Use existing infrastructure to manage data, resources and to collaborate to solve a large scale problem. – Use multiple computer (resources) to build a Virtual computer architecture – Open unlike Sync (closed)

Programming Model (Environment) • Scalability Criteria Systems Sync Co. Frame WS Lang to describe Service Java WSDL Coding Language Java Any Scalable No Yes – Firewalls • Description of Services: – Co. Frame: WSDL (Web Services Description Language) – Sync: Java • Coding Language – Co. Frame – Sync • Globus Toolkit 3 – Open. Source – Industry Standard

Programming Cost (Transparency) – Developers need to only concern themselves with services being offered and not how the services are implemented. – Co. Frame Architecture: • Grid layer separates Application layer from Resource layer • Younas/Iqbal doesn’t provide such an abstraction. • Sync (Integrated RMI)

Consistency of Data – Younis/Iqbal propose a transaction based model similar to ACID concepts – ACID (Atomic, Consistent, Isolated, Durable) – SACRe. D – Co. Frame doesn’t mention Consistency of data but the SACRe. D approach can be implemented in the CMB part of the framework Criteria Systems Sync Co. Frame WS Merging Support Yes NDE

Fault Tolerance Criteria Systems Architecture Sync Centralized Co. Frame WS Decentralized • Central server easier to secure (one point of failure) – Grid Architecture (Distributed) – Co. Frame provides for: • Single Sign On Service

Resource Optimization • Bottleneck : Network Bandwidth and Latency – Effects: • Deterioration of user experience – Example: • Video Chat with video frame updating every two seconds • How to Fix this: • Improve resources – Better network layout • Optimize current resources – Message Compression – Resource Replication Policy

Resource Optimization: “Message Compression” • Why Needed ? – Collaborative applications share a lot of data. – Inter-collaborative application messages – Example: • Dragging a window across screen – Updates need to be sent consistently to all connected clients – Use of XML – Message must include: • Sender id, message id, application id, timestamp, message type, fieldnames and data values corresponding to each field • Send messages less frequently ? • Encoding before sending, decoding on receiving – Gutwin proposes a mix of Ziv-lempel algorithm, Huffman Coding and building a dictionary of sequences with short lookup codes.

Message Compression • Sources of Inefficiency in Groupware messages – Repetition Within a single message – Repetition between messages – Inefficient Coding • Encoding of numbers. • Field widths for numeric data types.

Message Compression - Algorithm • Within-Message Issues: – Use standard zlib algorithm • Helpful when sending large messages • Between-messages Compression – Treat one message as template – Compare subsequent against that template, to determine repeated sequences – Replace repeated sequences by codes in a dictionary – Create new templates if messages not similar

Resource Replication • Replication of resource – Backup of Data – More optimized – can serve from best source • Centralization of resource – Easier to implement • Grids are distributed in nature – Replication model proposed by Co. Frame

Resource Replication • Resources stored independent of working nodes • Provider publishes to one server – Broadcast message using CMB to store metadata

Heterogeneous Systems • CSCW Application built for different resources – Example: Two collaborators (A & B) connected using two different devices – A has good Network (Cable/DSL) while B is on a relatively poor network (Dialup) – Both on a Video Chat or using a 3 D graphics application • Compression of data • Lower bit-rate

Conclusion Criteria Systems Sync Co. Frame WS - - - N Y N RMI C. REG. UDDI RMI/SOAP Programming Model (Environment) Central Dist. Programming Cost (Transparency) Y Y Y Consistency - - Merging Y ? ? Architecture (Centralized/Decentralized) - - - Resource Optimization - - Message Compression N NR Y Y Flexibility (Automation) - Awareness Support - Directory Service - Communication - Resource Replication Policy