19
1 EASY-R an SDMX Based Timeseries Repository Oliver Lorenz Swiss National Bank SDMX Global Conference, 2011

SNB‘s new SDMX based Timeseries Repository

  • Upload
    others

  • View
    13

  • Download
    0

Embed Size (px)

Citation preview

Page 1: SNB‘s new SDMX based Timeseries Repository

1

EASY-R an SDMX Based Timeseries Repository

Oliver Lorenz

Swiss National Bank

SDMX Global Conference, 2011

Page 2: SNB‘s new SDMX based Timeseries Repository

2

Workflow

banks

industry

external publications www.snb.ch

internal publications Intranet

internal data sources

investment funds

statistical surveys aggregates

• banking statistics

• balance of

payments

• investment fund

statistics

Primary

Statistical

System

Central

Timeseries

Repository

Page 3: SNB‘s new SDMX based Timeseries Repository

3

EASY-R

~ 5 x 106

timeseries units:

• Economic Analysis

• Inflation Forecasting

• Monetary Policy Analysis

• Financial Stability

• Financial Markets

200 users

• standard users (80%)

• expert users (20%)

• administrators (2%)

Excel

User Interfaces

• Data Viewer/Editor

• Structure Viewer/Editor

• Scripting Environment

• Rights Management

Page 4: SNB‘s new SDMX based Timeseries Repository

4

Life-Cycle Coverage

EASY-R

Analysis

Processing

Warehousing

Management

Dissemination

Page 5: SNB‘s new SDMX based Timeseries Repository

5

Several Generations of Home Built Systems

• SNB has a tradition of home built systems

• Tailoring “Workflows EASY”

• Competence and confidence to design

and implement the Next Generation

System EASY-R

EASY (mvs)

197x - 2000

EASY (vb)

2000 - 2008

EASY (.Net)

2008 - 2011

EASY-R

2010 – 202x

Page 6: SNB‘s new SDMX based Timeseries Repository

6

EASY-R Innovation Goals

User Interfaces for 3 Distinct User Groups

Process Oriented Repository

Domain Model

Uniform Technology Stack

Page 7: SNB‘s new SDMX based Timeseries Repository

7

Development Approach

7

EASY-R

Internal Domain Experts

Internal „timeseries systems“ Experts

In-House Developmen

t (with Contracted

Teams)

Agile Project Management

(SCRUM)

Open Source

Technology

Page 8: SNB‘s new SDMX based Timeseries Repository

8

Governing Principles

• Flexible Business

Domain Model defined by economists for economists

• Self-Descriptive Data with ‚proper/formal‘ semantics

• Metadata Based

Configuration of the repository, business logic and GUI

• Retrieval using the structures and semantics of the model

• Full Revision History to keep track & undo, revert, reconstruct all changes

• Workflow using configured staging to separate production phases

• Integrated Scripting for analysis, transformation and maintenance

Page 9: SNB‘s new SDMX based Timeseries Repository

9

EASY-R

Repository

Easy-R

Domain Model

SDMX-based

Time Series

Engine

Various UIs

Scripting

Engine

Conversion

&

Synchronization

Component Architecture

Page 10: SNB‘s new SDMX based Timeseries Repository

10

Flexible Business Domain Model

and Self-Descriptive Data

10

receive structure receive data &

metadata start working

meta

data/structure

based, automatic

configuration

Page 11: SNB‘s new SDMX based Timeseries Repository

11

Flexible Business Domain Model

and Self-Descriptive Data

SDMX based data exchange

SDMX based repository

SDMX based time series arithmetic

11

receive structure receive data &

metadata start working

Page 12: SNB‘s new SDMX based Timeseries Repository

12

SNB’s SDMX-compatible Extensions for a Data

Warehouse

SDMX based Data Exchange

SDMX based repository

SDMX based time series arithmetic

12

receive structure receive data &

metadata start working

• Repository

– store and manipulate

structure or series

– validate structures or series

– efficient storage model

– workflow and data history

– user and rights

management

• Time Series Engine

– calendar

– arithmetic

– dynamic series

Page 13: SNB‘s new SDMX based Timeseries Repository

13

Time Series View

Page 14: SNB‘s new SDMX based Timeseries Repository

14

Structure View

Page 15: SNB‘s new SDMX based Timeseries Repository

15

Time Line and Costs

2006 2007 2008 2009 2010 2011 2012 2013

Migration (Data & Processes)

Inception & Elaboration

Construction Production R1

Construction I&E Production R2

• Costs of Construction Release 1 (estimates)

• IT:

• contracted SW Architects and Engineers : ~15PY (5Persons x 3Years) ~ 3,3 Mio CHF

• internal SW Architects and Engineers: ~ 3PY

• Software: < 20 kCHF

• Business:

• Workshops (Concepts, Planning, Review): ~ 3PY (5P * 20% * 3Y)

• Costs of Migration (5 Mio TS & ~180 Processes from 10 Business Units) • IT:

• contracted Support: ~5PY (2P * 2,5Y) 1,1 Mio CHF

• internal Support: ~3PY

• Business:

• Training, Planning, Preparation, Implementation, Testing, Deployment: > 1.5 PY (30P * 15Days)

Page 16: SNB‘s new SDMX based Timeseries Repository

16

Conclusions today

• The innovative vision of the new system could be

implemented with the chosen technology and architecture

• SDMX helped a lot (with some overhead)

• There is a comprehensive set of high quality Open Source

products, tools and frameworks available

• internal domain experts and IT competence (consulting,

project management, time series based information systems)

was vital for the success of the project.

• The project would not have been possible with a constant

and close support/involvement of the economists (expert

users from Statistics and Economic Analysis) (countless

workshops, meetings for planning and review)

Page 17: SNB‘s new SDMX based Timeseries Repository

17

Lessons Learned: Migration

• shared responsibility between IT (sum and quality of functionality) and economists (correctness and timeliness of migrated business processes)

• generaly at least as expensive and complex as building the system (time, resources)

• costs and complexity significantly reduced with self built solution – close to the original algorithms and processes

– resources & know how for user training

– IT resources to prepare „initial migration“

– 1:1 migration of data with legacy keys

– transparent interface to read from the legacy system

– improvement and adaption of the system with input from migration tasks

• the longest delay in the project roadmap

Page 18: SNB‘s new SDMX based Timeseries Repository

18

Lessons Learned: SDMX - TCO

• cost to learn, implement, adopt

• cost to interpret and extend

• cost of training (developers, data administrators, expert

users)

• effort to follow (and influence) the development of the

standard

• SDMX is quite heavy at times

• costs to extend a standard meant for data transmission

for use in a repository context (structured identifiers,

revisions and stages, process-information, database-

mapping, CRUD-performance)

Page 19: SNB‘s new SDMX based Timeseries Repository

19

Lessons Learned: SDMX - ROI

• a standard -> common vocabulary

• an evolved information model -> head start for the project

• data comes with describing metadata -> self configuring system !! – no need to restructure or transform external data

– a feasible way to describe our own data

• one format for all systems (internal, domestic format) timeseries, graphic system, statistical system

• standard for data exchange with other institutions – guaranteed correctly formatted SDMX

– SDMX sent from outside is checked during import (with some tolerance for inconsistency)

• extensible through annotations

• being a good SDMX citizen