[gridcpr-wg] XCAT3: In scope or out?

["Paul Stodghill" <stodghil@cs.cornell.edu>]

XCAT3 \cite{xcat3} is a Common Component Architecture application
framework based on Grid standards. One of the functions that XCAT3
provides is checkpointing for CCA-based applications. Because these
applications can be executed on a number of distributed computing
resources, consistency is a consideration when checkpointing. XCAT3
handles this by providing \emph{Application Coordinators}. When a
checkpoint is required, the user or some other agent notifies the
Application Coordinator, which then executes a non-blocking coordination
protocol between the distributed components.

 

In order to provide checkpointing within a heterogeneous computing
environment, XCAT3 uses application-level checkpointing. Also,
checkpoint data is stored in XML to ensure maximum portability.

 

In order to ensure the availability of checkpoint data in the event of
processor failure, XCAT3 assumes a Storage Service Federation, which can
provide stable storage for checkpoint data.

 

Functional requirements:
\begin{itemize}
\item API for application state writing and reading.
\item Services for checkpoint data management and associated API.
\end{itemize}