
220 Practical Workflow Applications
into lists of files, the files are then given as input for parallel processing
by distinct nodes of a cluster. BeesyCluster offers the ability to restart
processing on other nodes in case some services have failed. This im-
plementation is sufficient if partitioning is a relatively quick process.
Otherwise, it is possible to use the streaming mode for configuration of
pipelining in the following way (Figure 7.25b):
(a) The service based on data
split would run in the streaming
mode as would the one based on process data packet multiple
options.
(b) data
split would partition into a larger number of smaller files
(with a smalle ...