Parallel computing strategies in the analysis of the inhibiting effect of price limits on futures prices

摘要

In futures markets with price limits, trading halts are triggered by limit hits. Limit hits are rarely observed,rnperhaps because traders avoid bid-ask quotes that cause them. If this explanation is true, futures prices wouldrncluster in a narrow region close to the limits. We test this empirically for currency futures contracts and findrnresults consistent with the explanation. The tests require calculations of all combinations of a computationallyrnintensive time series, which are extremely time consuming on a sequential machine and hence limit thernpracticality of the analysis. Consequently, we investigate parallel computing strategies in partitioning therndatasets and solving them in parallel on a high-end Beowulf cluster. We discuss two different partitioning strategiesrnof the given datasets on the cluster and elaborate the results.