At least put the AEs and the consumers into separate CAS processors. Otherwise you run into either of the two situations:
1) AEs are not parallelized - no performance gain 2) consumers are parallelized - depending on their implementation, you might be getting mangled data, missing data, etc. Cheers, -- Richard > On 06.09.2016, at 10:56, armin.weg...@bka.bund.de wrote: > > Hi! > > What's the best practice to combine analysis engines into CAS processors? > Should every analysis engine become its own CAS processor? Should analysis > engines be combined to aggregates which become CAS processors? What are the > conditions for doing so: technical, semantical, logical? > > Best, > Armin >
