Has there been much written about how UIMA-AS has been deployed?
Would anyone care to take a moment to describe their experience?
Do you typically create a short main pipeline that delegates out to
many instances of a longer one? I imagine it would depend on the
load characteristics of each engine, but I could see a collection reader
and a pipeline that delegates to an aggregate that has all the analysis
engines in it. At what point do the central pipeline, collection reader and
broker saturate that machine?
thanks
-Chris
Jaroslaw Cwiklik wrote:
> Chuck, good news! Just FYI, while you are trying the CPE please keep in mind
> that there is a new generation scaleout framework for UIMA called UIMA AS.
> It is based on JMS and ActiveMQ specifically. It scales out very nicely over
> hundreds
> of compute nodes. You can read about it here:
>
> http://uima.apache.org/doc-uimaas-what.html
>
> If you are planning new development with UIMA I would advise you to try the
> UIMA AS as you scaleout platform for UIMA.
>
> Regards, Jerry
>
> On Thu, Apr 22, 2010 at 1:06 PM, Chuck Bearden <cfbearden@gmail.com> wrote:
>
>
>
|