aldorexbraam

My take-aways from Vcampus Live! 2013

Discussion created by aldorexbraam on Dec 11, 2013
Latest reply on Dec 16, 2013 by lgarrigues

To start with, I really liked to be there on vcl13.

With this post I want to share my personal opinions and take-aways from vcl13 with the community with the objective to start up some interesting exchanges of thoughts.

The Good

PI ODATA services: This is a great leap to offer business analytics. Clients who take the effort of investing in a structured asset model on top their PI data will greatly benefit from this.

 

PI Core Webservice: This has been coming for quite a while and I like the idea of cutting lose from the need to install bulky SDK’s to get to my PI data. I perfectly caters for the mobile/mashup case.

 

Esri maps integration with PI data( Feature services): I really liked the idea/concept of ‘pushing’ your geomatics + real-time information to the feature services and getting served full featured maps, well done.  Blind spot for me is data protection and security.

 

EventFrame Generator: personal interest having written by own eventframe generator because the PI Notifications 2012 functionality did not have capability to generate eventframes nor (more importantly) do this is retrospective. I liked to see this requirement covered and fully integrated in the product stack, especially in combination with asset based analytics.

 

 Asset Based Analytics: I am in love with this……ABAcus totally rocks!! The ability to use PE’s, but more in a divide-and-conquer way, in small steps make the PE’s much more readable/maintainable, also in connection with the AF asset model attributes (which might be formulas). The backfilling option also rocks and meets a business requirement I am currently dealing with where we are ‘tuning’ emissions estimations and need to redo calculations once we change parameters…not sure how it will work out when you also want to recalc a totalizer….need to dive in the manuals for that

The Bad

PI Core Webservices:  credited this was a beta version, I found the support of bulk fetches a bit annoying. Simple things like retrieving current values from attributes require lots of coded calls to the rest API. I would have like the REST api to be a bit more asset model ‘aware’ in the sense that you can pass in the element path or construct a list of element paths like the AFAttributeList.GetValue method.

The ugly

PI manual logger 2012: Oh man…this tool reeks of its history…I would not approach my customers with this…they will kill me..Plain ugly windows 3.1 style interface, no integrated security, no decent reporting capabilities except for tapping directly into the database…

The puzzling

PI Cloud Services: The PI cloud services is seemingly a compelling story, but I am uncertain of the actual value proposition in this case. What are say Oil majors, Petro chemicals and Pharma companies getting out of this? Most of the companies are very secretive about their production data, aside from the fact that in some countries there are even laws stating that production data is not permitted to be hosted out-of-country. Data that these companies DO want to expose is I think of a different time dimension, mostly aggregations. And for this there are currently already other cloud solutions around.

 

StreamInsight : where is it ? The silence on this is deafening ….I know this is more a Microsoft ‘thingy’, but it was supposed to be a first OSI/M$ banding together….I looks to be born dead…not sure what the alternatives are….

 

 

The Promising

I attended the VOX pop on the Datawatch product…perhaps I was a bit to direct in my comments, apologies for that ( but heck, I am Dutch) . I actually liked the visualizations…I  got me thinking about some fundamental questions:

 

1)      For BI purposes, Can we filter/do analytics directly at the event pipe, in order to do data cleansing ‘at the gate’ instead of post mortem ?

 

2)      If OSIsoft is really serious about real-time BI, where the challenge also lies in preventing the ‘data copying’ . I would expect them to work on an adapter that taps directly into the raw archive system and  where  you can ‘enrich’ this information with (meta) data from your asset model.

 

3)      Related to (1). Would it not be great to have a sort of Map/Reduce functionality inside/close to the pi server where you could plug-in your (predictive) analytics logic ?

Outcomes