Transcript CCF: Collaborative Computing Frameworks

CCF: Collaborative Computing
Frameworks
Vaidy Sunderam
Emory University, Atlanta, USA
NSF Multidisciplinary Challenges Research Project - ASC 9527186
 Overview
–
–
–
Internet Collaboratory
Integrated framework
for collaboration, data
mgmt, computation &
communication
Virtual interaction and
work environment
CCF: Project Overview
 Goals
– Virtual Work Environment
– Internet based Collaboratory
 Research
– Internet/2 Distributed Computing
– Collaboration Methodologies
– Integrated computation/communication/data
management
CCF Operational Scenario
 Participants
WS Cluster
Biochemist
Chem
Lab LAN
NMR
Internet
links
initiate or
join a “session”
 Cooperation via
–
–
–
Digitized data
–
–
Parallel processor
Pharmacologist
shared virtual desktop
audio/clearboard/chat
shared data space
computational tools
[communications
fabric]
CCF Elements
 Communication
 Shared
workspace (display, clearboard)
 Computation
 Data
infrastructure (SW)
tools/resources
management (object sharing,
manipulation, referral, archiving)
 Telepresence/interaction (A/V/text)
CCF Virtual Desktop
 Multi-display,
shared
output/graphics
 Multilayer annotation
and telepointing(CB)
 Ordered text (Chat)
and mixing f/d audio
 Data space, compute
tools, launching apps
Virtual Desktop - CCSM
 Components
– X Windows based, 1/8/24 bit
– Menus for applications,tools,session mgmt
 Unique
features
– WYSIWIS area, “grayed” detachment
– Scalable, minimal state, failure resilient
– Free/controlled input and layout
– Heterogeneity support
CCFX and CCSM
 CCFX
– Emulates very flexible and general X server
– Multiplex clients, multicast protocols
– CCSM: Interaction GUI, callbacks, window
management, desktop/viewport
 Technical
Issues
– Visuals (8/24 color), keyboard, multiple conns
– Heterogeneity, WYSIWIS, floor control, perf
– Latecomers
X Sharing - CCFX and IMUX
X
Multiplexing
– Client <->generalized
pseudo-server(CCFX)
– Multicast to CCSM’s

Input Multiplexing
– Replicated clients
– One master; input
sequences duplicated
CCF Tools
 Audio
–
–
conferencing
Many users, realtime software mixing, silence
detection, handsfree, full duplex
GSM or ulaw, anti-clipping, control GUI
 Text-based
–
–
multiway chat
Atomic multicast: fully ordered wrt all
participants (messages user-tagged)
Single or multi-line input, scrolling, archival
and logging capability
CCF Clearboard
 Transparent
overlay
– Annotations of
dynamic entities
– Object-oriented marks
– Collaborative mark
manipulation/label
– Concurrency control,
latecomers, “base win”
– Full toolset
CCF System Architecture
Applications
 SW
–
CCF Tools
Chat
CB
Audio
–
?
SW
DSM
Virtual Desktop
CCFDS
CCFCF
CCSM
–
CCFX
CCF(naming, white pages, uniform API)
CCTL(multiway transport groups QoS)
Internet(UDP,IP mcast), ATM etc

Infrastructure
CCTL transport layer
Data management &
computing frameworks
Graphical interface
End applications
–
–
–
Collaboration tools
User applications
Extensions and plugins
CCTL - Communication Layer
Application
Channel
Comms.
multimodal
transport protocol
–
CCTL API
Session,
total ord
 Multiway
Channel
Membership
–
UDP, IP mcast, DLL
–
Networks
–
1 to 1, 1 to N, N to N
exchange
sessions (heavyweight
groups) and channels
(virtual synchrony)
multiple QoS:unreliable,
atomic mcast, ordered
low latency, high
bandwidth transport
CCF Data & Computation
 Shared
–
–
virtual data space
Typed objects: files, fifos, scripts, filters
Objects are imported (and un-imported) into
CCFDS via projections of underlying objects;
concurrency control, ownership, security
 Heterogeneous
–
–
network computing
parallel/distributed processing on VM
Upload/initialize agents/applications; message
passing interaction, migration
CCDS - Data Space
 Shared
Server
Client
Client
Client
Virtual
Storage
Server
Server
data space
– simultaneous access
with multiple levels of
concurrency control
– virtual object types:
files, filters, active objs
– latecomer support for
collaborative sessions
– data management tools
and utilities
CCF Computation Framework
 Heterogeneous
network computing
– parallel/distributed processing
– harness aggregate compute power & special
capabilities
– confluence:compute+data+viz+invoker
 Architecture/Model
– VM: host pool+external (free/constrained)
– Upload/initialize agents and applications
– Message passing interaction: entities - invokers
facilitators, compute modules, data, visualizer
IceT: Heterogeneous Computing
 Concurrent
Processes Processes
User
Data
0101
1011
1000
0101
1011
1000
Data
Computational
Resources
User
Computational
Resources
Data Processes
0101
1011
1000
User
Computational
Resources
Computing
– merging and splitting of
multiple virtual machines
– portability of code and
data, uploading, soft
install and migration
– Java, Java-C, C, and
Fortran support
– message passing parallel
and distributed computing
Extensibility, Potential
 Framework
– Encompasses essential elements of
collaborative computing in layered architecture
– Software backplane for collaboration
components; well defined interfaces
 Applications
–
–
–
–
Scientific: computation, data, interaction
Commercial: databases, financial, conferencing
Defense: distributed simulation, C3I
Education: Distance learning, virtual classroom
CCF: Summary and Status
 Integrated
–
–
Framework for Collaboration
Interaction, Communication, Data, Computing
Scalable, stateless, robust, extensible
 Research
Software
– Future work: multiple platforms, consolidation
– Alpha release available
– http://ccf.mathcs.emory.edu/
–
Contact - Vaidy Sunderam, [email protected]