Retries. When a workflow has a few input files (it has a few iterations) if something goes wrong or one of this files fails at some point of the workflow there is always the option of running the workflow again. However when there are a lot of iterations running the workflow again is a waste of time and resources. Taverna implements an automatic retry system that allows the designer to configure every web service call in a workflow. Figure 4: retry parameters In Figure 4 it can be seen how to configure retries for a web service and the parameters involved. The Taverna tutorial has a link to a myGrid video which is very descriptive and helpful to understand how to use the “retry system” in Taverna.
Retries. Defines the maximum number of http requests that a crawler will make in order to fetch a web document. If there is no response from the server after Retries times, the web document will be skipped. The default value is 10. This option is disabled for users accessing the service via the web application.
Retries. The Retries parameter designates the number of times information will be retransmitted when an acknowledgement is not received. The default setting is 3 times. Optional IP Settings Figure 14 shows a screenshot of the optional IP settings.
Retries. In the previous experiments, to better control the rate of cross-shard transactions, clients submit transactions only if they know the contracts are not going to be moved when the transaction is executed. To better model transactions in the presence of conflicts we experiment with the SCoin contract without any help from external sources. Two situa- tions can make clients retry transactions: when performing a single-shard transaction and the interacted contract is moved to another shard or when performing a cross-shard transaction and the called contract is moved to another shard. In the experiment we make clients wait a random time corresponding to the creation of 0 to 10 blocks if the transaction fails for any of these reasons, this is done to prevent contracts moving back and forth in an endless cycle. Figure 7 (right) shows the ideal latency where no conflicts exist, and Figure 7 (left) shows the latency when conflicts can happen. A clear increase in latency is observed when comparing both figures, but when retries can happen the number of times the same transaction fails and has to retry is highly skewed, for instance, 66% of the transactions that retry, do it just once, and only 1% of these transactions are retried more than three times.
