An InfoSphere Streams based approach for implementing an FX-style auto-correlation spectrometer with accelerator support
暂无分享,去创建一个
e signi cantly reduced data is persisted to a suitable store by this SINK PE. e persisted data can be used to perform longer integrations or deeper spectral analysis. UDOP (User De ned OPerator) PEs are used to implement the accelerator device speci c logic on the respective host. e accelerator devices are used to perform the time to frequency domain conversion. Communication between host and accelerator device is performed asynchronously in combination with multi-bu ering. In most cases the initial size chosen for a tuple to be optimal for a particular type of accelerator device usually implies that it contains data values for many frequency domain operations. Hence this FUNCTOR PE performs an integration operation to aggregate the auto-correlated results for the frequency domain operations performed within a tuple. SPADE Application Data ow Graph
[1] Chun Zhang,et al. Storing and querying ordered XML using a relational database system , 2002, SIGMOD '02.
[2] Philip S. Yu,et al. SPADE: the system s declarative stream processing engine , 2008, SIGMOD Conference.