BTW by default the EventServer is on 7070 and the PredictionServer for a 
specific engine is on 8000


On Sep 30, 2016, at 1:12 PM, Pat Ferrel <[email protected]> wrote:

The two major components of PIO are the EventServer, which is where you send 
events, and the PredictionServer, which is where you send queries. An engine is 
a specific template, which supplies the heart of the PredictionServer as well 
as doing model building with `pio train`

Some workflow diagrams are here: http://actionml.com/docs/pio_architecture. 
With PIO you input events, periodically train to build a model, and serve 
queries. The internals of these for the UR are described on the page mentioned.

Spray is a REST HTTP platform used to put up REST interfaces. We are replacing 
it with the latest evolution of that lib, now called Akka-HTTP

On Sep 30, 2016, at 5:55 AM, Gustavo Frederico 
<[email protected]> wrote:

Hi, I installed PIO+UR with the default settings. I see the process
with arguments

...  --engine-variant file:/home/mengao/timao/engine.json --ip 0.0.0.0
--port 8000 --event-server-ip 0.0.0.0 --event-server-port 7070 ...

A basic question: what's the difference between the 'engine' and the
'event server'? I see the engine port 8000 and event server port 7070.

Also, can someone please explain in general terms "a day in the life
of a request and a response"? That is, how the data travels among the
major frameworks ( Event Server, Elasticsearch, Sparks, HBase, ... ) -
assuming Elasticsearch + Hbase and only 1 server

Footnote: what's "spray-can" ? I see "Server: spray-can/1.3.3" as a
header in one http response.

Thanks!

Gustavo


Reply via email to