2017-05-16 Meeting notes
- Dennis Wisnosky
Date
Attendees
Agenda
1) Use Case reminder
2) Where we are on our road map.
3) Open Action Items
4) JIRA Issues Review - https://jira.edmcouncil.org/projects/FCP/issues/FCP-4?filter=allopenissues
5) Todays content discussion.
SMIF OWL-UML
SKOS
RDF/S
6) For next week.
Proceedings:
20170516 FIBO CFTC PoC
JN: Need to show the advantages to the business. Scrini likes the triple store. DN: What did he like? It enables the silos to be torn down by connecting data. Not using 1982 technology like today. DN: This sounds like smart aggregation is what he likes. JN: Right, exactly!
JN: Need to be able to use SAS for analysis also. DA: There is GItHub open source project that connects SAS to triple stores. DN recalls the same, or through CSV's.
http://www.phusewiki.org/wiki/images/f/f2/Semantics101-PhUSE-EU-CSS -Pt4.pdf
https://github.com/MarcJAndersen/SAS-SPARQLwrapper
JN: SDR data is inconstant and needs to be fixed by CFTC. FIBO can deal with this. Scrini liked this also. And, said this will take some work with the SDR's. Wanted to know how change is managed. Much easier than with current relational data bases. For the presentation we should show what we can do with the TS.
DN: We should do whatever it takes to cross the Rubicon. Do onto and then TS easy. Scrini wants TS and then show ontology. MA: Sell the sizzle. Max: a triple store is just a rudimentary ontology.
JN: Need to have a clear way to show how changes will be managed. We have already done TS and can prove that it works. Just want to do more now. Need to show that the scalability is through FIBO to connect the triple stores. Then can pull in data from anywhere anytime with total flexibility and scalability. Performance is not an issue MarkLogic has proven better performance than with relational data stores. Scrini was just using Socratic approach. JN: Good to have the FIBO experts as support.
MA: Now in discussions with ISDA and DTCC that directly relates to this CFTC work. JN: Good news!
Scrni wants to see the demo b 4 John Rogers sees it.
DN wants a slide reviewed and changed. DN shows a Protege page that JN says is perfect. DN will send JN a screen shot of this.
JN: This shows exactly how QA and QC validation of the data can be done. Dean shows TS first and what can be done and then we show what is behind the tool to make sure the data is good. Much better than spread sheets and other dumb approaches. Max: Any data base can be expressed as a TS and then it is an ontology and more useful and can add more relationships and more definitions. All technology exists to do all of the queries.
DN: Leveraging the inherent nature of ontologies for data aggregation should be enough to win the day. JN: I did not make the case well enough. Need FIBO experts to opine on all of this. You guys have more weight.
DN: What is expectation on the outcome tomorrow? What should we drive for, or hope for? JN: Get buy in. There was a question about cost. I said not much because this will be small, not enterprise wide at first. Need $ for tools (TQ) and a TS and consultants. It would be good if we have a real number.
MA: What would be of most use to CFTC. JN: Full interest Rate Swaps. Probably can do this for $100,000. Not a big deal for CFTC. Tend to think in terms of millions. Bring in consultants under NDE. Then EDMC add what is needed to FIBO.
Max, the information content is what drives the cost.
JN: Could give $$$ to EDMC, but probably not fly. Should discuss. Could give $$ to consultants who work for CFTC & are FIBO insiders. We need to throw out a number. What would IRS cost. DA: Depends on the deployment. JN: Data from all of the SDR's delivering CVS data. DA: What to do with the data? Dash boards or integrated to EDM systems. Stand alone much easier.
JN: 4 SDR's into triple store to do analytics and put into SAS. That is it.
DN: Need to analyze the complexity of the data to see how much work needs to be done. Need to determine how many fields from the SDRs are necessary so that a real estimate can be made. DA: FIBO has a model of the Swaps augmented by SS PoCs. May still need more. DN: 4 SDRs great, but need to clearly define scope of the data that will be used. The kind of Swaps for example need to decide and then defer work with other data till later. Classify the types of IRS's and then get the data and then do analytics.
JN: Just do IRS as one of the asset classes. Most of the analytics are 42 fields. Collect lot more, but use 42. Then could do e.g. GS exposure to Treasury Bonds. GS has 3 data bases that don't talk to each other.
MA: Full range of IRS' linked to the SDR's mapped to FIBO. Not a big deal. DN: Yes, not a big deal. But, Max points out that FIBO exposes the interoperability of financial instruments. Never done before. Can't be done any other way. Shows why FIBO is the direction for the future. DA: Demo makes this point. We could do an interesting project as described for somewhere around the $$$100 JN said.
JN: 42 fields for IRS analytics. The 42 include many identifiers. I want to tell them, "roughly $100K. One million would lose them. DA: Million necessary for enterprise level. DN: not less than $100K, not greater than $250K. Highly variable depending on #'s of people involved.
JN: More than a PoC, but less that full implementation. Handful of licenses. MA: IRS test case using live data. JN: Like saying a test case.
MA: Say $100K to $200K. JN: Sticker shock is anything greater than $100K. JN: Not test case - Limited Implementation. DN: Need to know what the analytics are so we know what is connected, i.e. Anzo, R, or what.
DA: Yes, small step first. DN: Anzo would be a good start because we know how to use it. Need good price for Anzo.
DA updated his slide and showing it now. JN likes this.
TSD is transitional swaps data with 42 fields. SDR is Swap Data Repository. ISS is position data for futures. TCR is Trade capture report, i.e. transactional data. Dashed line means not all data is in a single TS.
JN likes with DA did. David runs through the deck.
JN will update the next steps and we are good to go. JN: Will take just a few min on all of his slides to set the stage for DA.
JN: Rem there will be no decision made. Just need to get buy in by Scrini. CIO will make a decision when we meet with him.
If we are asked about cost, I will just say about $100K. Need to prepare a real proposal. JN: No, I will not say a cost at all. Will do this later off of the call.
If Scrini asks us, we say in the low $100's subject to scoping.