Consistency Control in Distributed Collaboration ToolsCISMM: Computer Integrated Systems for Microscopy and Manipulation

Project Leads: Diane Sonnenwald* and Mary Whitton Investigators: Tom Hudson+, Kevin Jeffay (UNC-CS), Russ Taylor * Now at University of Borås, Sweden + Now at the University of North Carolina at Wilmington

http://cs.unc.edu/Research/nano/

November 2003

The Problem: Provide application consistency AND interactivity The Solution: Use optimistic techniques wherever possible

Implementing Optimistic Concurrency Control in the Collaborative nM

Funding for this project was provided by NIH National Center for Research Resources

The nanoManipulator Collaboratory enables non-collocated users to work together to perform real-time experiments and/or analyze previously collected data from the Atomic Force Microscope (AFM). To ensure user acceptance, it had to have the same functionality as the single user system and maintain the same high level of interactivity.

In this system the application is replicated at each site and the network carries control signals, video, and application data between sites.

Concurrency Technique Suitable Usage Scenarios

Explicit Locks Not losing work is critical, even at the cost of latency and a changed workflow

Implicit Locks Not interrupting natural workflow is critical, even at the cost of latency or lost work

Coarse-Grained Locks

Objects to be acted upon are interdependent, either algorithmically or in users’ planning.

Fine-Grained Locks

Objects to be acted upon can be separated into independent subsets

Optimism As for fine-grained, implicit locks; ideally, conflicting user actions can be automatically reconciled; actions have little/no risk of damage (to devices or data) and are easyto undo

Techniques for consistency control which uselocks require one or more network round-trips before parameters change at any collaborator’ssite. Locks insure application state consistency and prevent damage to data and/or devices.

Optimistic techniques allow parameter changesto occur immediately at the local site, maintaining interactivity. When the change is propagated to remote sites, they either implement or ignore it depending on intervening events that happened at those remote sites.

Optimistic techniques maintain sufficient consistency for manipulations of individual visualization parameters; we use explicit locks to preserve users’ intent across sequences of manipulations of the microscope and prevent damaging the device.

To identify the appropriate concurrency technique for each shared item of data in our system, we used parallel and hierarchical applications of the Model-View-Controller paradigm.

The figure above shows two Model-View-Controller triples capture the different semantics of the microscope state and visualization parameters: microscope state must be pessimistically shared, while visualization parameters can be shared optimistically. The 3D rendering draws from both Models to provide an additional View.

Data with different consistency requirements was grouped into different models, while data that dealt with coherent units of program function was grouped into submodels within the model. We could vary the coupling and concurrency control of each model and submodel as necessary, giving us a very flexible infrastructure for experimenting with alternate implementation approaches and user interfaces for the Collaborative nanoManipulator.

Viewing Parameter Controls (submodel)

Replay Controls (submodel)Elapsed time in replayReplay rate

Color Controls (submodel)Source plane for color valuesColormap

Contour Map Controls (submodel)

Lighting Controls (submodel)

Visualization Model Rendered 3D Graphics View

Microscope Controller

Microscope Views

VisualizationController

Visualization Model(Optimistic Concurrency

Control)

Visualization Views

Microscope Model (Floor Control;

Coarse-Grained Locking)

PUBLIC NETWORK LAN LAN

PC-1: Shared nM application- AFM control- visualization

Phantom

PC-2:Net Meeting: - video conf.- shared apps.

Phantom

PC

AFM

User at Location A User at Location B

PC-1: Shared nM application- AFM control- visualization

PC-2: Net Meeting: - video conf.- shared apps.