Schema On Read In Financial Services Presented by: Lee Pollington, Solutions Director, Amir Halfon, Financial Services CTO © COPYRIGHT 2014 MARKLOGIC CORPORATION. ALL RIGHTS RESERVED. Financial Services Use Cases SLIDE: 2 Trade Operational Data Store Reference Data Management Regulatory and Legal Compliance Customer Insight Financial Crime Prevention Information Distribution © COPYRIGHT 2014 MARKLOGIC CORPORATION. ALL RIGHTS RESERVED. Old Generation Trade Store: Schema on Write Limited, fragmented analytics and reporting capabilities Expensive, error-prone post-trade processing Long, costly development cycles Derivatives Rates FX Matching ETL … Clearing Settlement …etc. …etc. Multiple Relational Data Stores for different instrument types SLIDE: 3 © COPYRIGHT 2014 MARKLOGIC CORPORATION. ALL RIGHTS RESERVED. What’s the problem Huge Volumes of Data – (billions of trades per day, PBs/TBs per day) 10s of Source Systems 10s of Users of Different Roles Few Source Systems + Single Client Cross asset view One schema can’t work There are two time axes, valid & system SLIDE: 4 ✔ ✖ ✖ © COPYRIGHT 2014 MARKLOGIC CORPORATION. ALL RIGHTS RESERVED. The Back Office Landscape Fraud Detection 10s Analyst Derivatives Securities Uncleared Margin Reporting Accountant FX Liquidity Manager 10s Reference Data Liquidity Reporting Reference Data Reference Data SLIDE: 5 © COPYRIGHT 2014 MARKLOGIC CORPORATION. ALL RIGHTS RESERVED. The Mechanics Of The Schema Problem 35667 75335 95577 TRADER_ID T_ID TID James Gold Ted Cash Amy Wedge AN DT PB TRD_ID SLIDE: 6 TRID TRADERID 245 | AMF 980 | GHN 234 | QTE © COPYRIGHT 2014 MARKLOGIC CORPORATION. ALL RIGHTS RESERVED. Transformation, Mastering & Science SLIDE: 7 © COPYRIGHT 2014 MARKLOGIC CORPORATION. ALL RIGHTS RESERVED. Now Add Multiple Reference Data Systems Reference data describes traded instruments (stock, bonds) and the related legal entities (e.g. issuer, counterparty). Data has a temporal aspect (two time axes, valid and system) Corporate actions - companies go through mergers, acquisitions and spinoffs Local Reference Data Tactical Reference Data Corrections over time Provenance is critical (regulatory compliance) Strategic Reference Data SLIDE: 8 © COPYRIGHT 2014 MARKLOGIC CORPORATION. ALL RIGHTS RESERVED. Now Add Reference Data… Data Acquisition Content Management Data Distribution Trading Risk Golden Copy ETL ETL normalize, arbitrate … etc. SLIDE: 9 Raw Data (staging) Compliance Research etc. © COPYRIGHT 2014 MARKLOGIC CORPORATION. ALL RIGHTS RESERVED. What Are The Client’s Needs? Only the set of data I am interested in As fresh as possible In the format I understand In time chunks I can process With the ability to do bitemporal operations SLIDE: 10 © COPYRIGHT 2014 MARKLOGIC CORPORATION. ALL RIGHTS RESERVED. Schema on Read with MarkLogic Analyst Derivatives Securities Accountant FX Liquidity Manager Reference Data Reference Data Reference Data SLIDE: 11 © COPYRIGHT 2014 MARKLOGIC CORPORATION. ALL RIGHTS RESERVED. A Future Approach – Enter Semantics Provenance captured right along with the data Attributes were captured in an ontology Dataset mapping modeled ontology Semantic-based data integration SLIDE: 12 Dynamic inference at runtime to avoid “schema on write” © COPYRIGHT 2014 MARKLOGIC CORPORATION. ALL RIGHTS RESERVED. Data Provenance Using Semantics <Trade> <Cashflows> <PartyIdentifier> <TradeID>123456 </TradeID> </PartyIdentifier> </Cashflows> <provenance> <triple> <subject> Cashflows</subject> <predicate> wasDerivedFrom </predicate> <object> CDS_xyz </object> </triple> <triple> <subject> TradeID </subject> <predicate> wasAttributedTo </predicate> <object> System_123 </object> </triple> </provenance> </Trade> SLIDE: 13 © COPYRIGHT 2014 MARKLOGIC CORPORATION. ALL RIGHTS RESERVED. Reference Data in RDF Swap123 Leg1 JPMC counterParty hasLeg Cashflow Payment Date SLIDE: 14 Payment Amount Notional Currency BrokerDealer Derivatives LEI Equities © COPYRIGHT 2014 MARKLOGIC CORPORATION. ALL RIGHTS RESERVED. Semantic Data Integration and MDM Source systems ontology Source systems MDM … SLIDE: 15 Master Data Management 1. 2. 3. Match on metadata (table name etc.) Match on content (row count, columns, etc.) Determine action based on weighting: a. Auto-match b. Non-match c. Human intervention (workflow) Operations workflow © COPYRIGHT 2014 MARKLOGIC CORPORATION. ALL RIGHTS RESERVED. New Generation Data Management: Schema on Read Single store for all instrument types persisted as-is off a message bus Single source of truth for compliance Simplified workflow architecture Minimized trade exceptions Cost effective platform Matching Clearing Settlement …etc. Post Trade Processing Executed Trades Exceptions Management Surveillance, Risk & Compliance Load data “as is” From multiple refdata vendor feeds SLIDE: 16 etc. Deliver data to each consumer in the right format Historical Analysis © COPYRIGHT 2014 MARKLOGIC CORPORATION. ALL RIGHTS RESERVED. Recap – the Key Ingredients Document orientation Flexible Indexing Dynamic Transformtion Semantics Bi-temporality SLIDE: 17 © COPYRIGHT 2014 MARKLOGIC CORPORATION. ALL RIGHTS RESERVED. Q&A SLIDE: 18 © COPYRIGHT 2014 MARKLOGIC CORPORATION. ALL RIGHTS RESERVED.
© Copyright 2026 Paperzz