E-Business Suite Big Data Purge An Approach to Archive and Purge Financial Accounting Hub Data
Introductions
George Somogyi
IT Director
Societe Generale (Newedge)
Rey Mendez
Director, Oracle Applications
Avout Corporation
• Avout was formed in 2010 to
change the way consulting
companies delivered services
• Providing the Right Resource, at
the Right Time, at the Right Price
• Our Clients matter, our
Consultants matter, our
Executives matter, and the People
in our immediate and extended
community matter
• A world-leading multi-asset
derivatives broker
• 100% owned by Societe Generale
• Listed and OTC Clearing
• Agency Execution
• Prime Brokerage, incl. cross-
margining, financing, capital
introduction, research
• Access to 85 markets
2
Oracle Project Background
• Global Deployment Timeline
– North America – April 2011
– Canada – June 2011
– UK – November 2011
– France – January 2012
• Software Application Landscape
– Oracle e-Business Suite – R12.1.2
• Oracle Financial Accounting Hub, Oracle General Ledger, Oracle Payables,
Oracle Receivables, Oracle Assets and E-business Tax, Oracle Advanced
Global Intercompany Systems
– Hyperion HFM / FDM 11.1.2.3
– Reporting Toolset
• Essbase 11.1.2.3
• OBIEE 11.1.1.7
3
Fixed Assets (FA)
Advanced Global
Intercompany
System (AGIS)
TCA Payables
(AP)
Accounts
Receivable
Financial Application Architecture
Enterprise Reporting – OBIEE (Answers, Dashboards etc.)
NCR
Inte
gra
ted
Data
La
yer
Finance Data
Mart (FDM)
HFM
Essbase
Client Referential
Account Balances
and Shareholder
References
Account
Balances
COA Segments
Journal Header & Lines
Balances
General Ledger
Local GAAP, IFRS Ledger Options
Intercompany Detailed Journals
Financial Accounting Hub (FAH)
Subledger Accounting
Staging
Tables Pre-processing
Rules Engine
Net Business
Income
F10s
Futures
Balance Sheet
Trial Balance
Newedge Integrated
Data Layer
F10 Files
Ledger Options
Detailed Journals
Local GAAP, IFRS
Intercompany
Supporting Ref. w/balanceUser Tran. Identifiers
Rules Engine
Repository
Financial Accounting Hub / Subledger
Accounting
Payables
(AP)
Fixed Assets
(FA)TCA
Suppliers
Invoices,
Payments
Retirements
Depreciations
Transfers
Trading
Journals
Summary Journals
(IFRS, Local GAAPs)
Local CurrenciesDrill Back
General Ledger
COA Segments
Journal Header & Lines
Balances
EBTax
Advanced Global
Intercompany System
(AGIS)
Intercompany
TransactionsIntercompany
Payables, Receivables Journals
Finance Data
Mart
(FDM)
F10 information,
Oracle Code
Combinations
HFM
Account Balances
And
Supporting ref.
Balances
Enterprise Reporting - OBIEE (Answers, Dashboards etc.)
Pass Throughs
NCR
Suppliers
Procurement
Systems
Pass Through Trial
Balances
Oracle EBS Modules
Trading Source Systems /
Other Newedge Systems
Consolidation and Reporting Systems
Standard Product Interfaces
Custom Interfaces
Other Integration Points
Key
Asset
Systems
Legacy Trading Systems
Web ADI
Excel Uploads
Legacy
Lookups /
Referential
Manual Journal
Excel Upload
Essbase
Account Balances and Supporting
Balances
Manual Adjustment to only One Ledger
1
21
53
41
Concur (T&E)
ADP
Client Referential
Pre-processingStaging
Tables
Supporting Ref. w/o balance
PO Matching
T&E Invoice (approved
and paid)
Payment Information
Assets
(Conversion)
eUbix
Invoices
InvoicesTax
Engine
Net Business
Income
F10s
Futures
Balance Sheet
Trial Balance
Newedge Integrated
Data Layer
F10 Files
Ledger Options
Detailed Journals
Local GAAP, IFRS
Intercompany
Supporting Ref. w/balanceUser Tran. Identifiers
Rules Engine
Repository
Financial Accounting Hub / Subledger
Accounting
Payables
(AP)
Fixed Assets
(FA)TCA
Suppliers
Invoices,
Payments
Retirements
Depreciations
Transfers
Trading
Journals
Summary Journals
(IFRS, Local GAAPs)
Local CurrenciesDrill Back
General Ledger
COA Segments
Journal Header & Lines
Balances
EBTax
Advanced Global
Intercompany System
(AGIS)
Intercompany
TransactionsIntercompany
Payables, Receivables Journals
Finance Data
Mart
(FDM)
F10 information,
Oracle Code
Combinations
HFM
Account Balances
And
Supporting ref.
Balances
Enterprise Reporting - OBIEE (Answers, Dashboards etc.)
Pass Throughs
NCR
Suppliers
Procurement
Systems
Pass Through Trial
Balances
Oracle EBS Modules
Trading Source Systems /
Other Newedge Systems
Consolidation and Reporting Systems
Standard Product Interfaces
Custom Interfaces
Other Integration Points
Key
Asset
Systems
Legacy Trading Systems
Web ADI
Excel Uploads
Legacy
Lookups /
Referential
Manual Journal
Excel Upload
Essbase
Account Balances and Supporting
Balances
Manual Adjustment to only One Ledger
1
21
53
41
Concur (T&E)
ADP
Client Referential
Pre-processingStaging
Tables
Supporting Ref. w/o balance
PO Matching
T&E Invoice (approved
and paid)
Payment Information
Assets
(Conversion)
eUbix
Invoices
InvoicesTax
Engine
Net Business
Income
Futures
Balance Sheet
Trial Balance
Legacy Trading Systems
Pass Throughs
F10s
Web ADI
Excel Manual
Journal Uploads
4
EBS Business Landscape
Ledgers • North America
• US GAAP, primary ledger, ledger 2024
• US IFRS, secondary ledger, ledger 2025
• Canada, primary ledger, ledger 2028
• UK: IFRS primary ledger
• French
• FR GAAP, primary ledger, ledger 2053
• FR IFRS, secondary ledger, ledger 2091
Reporting Components
• One Global Chart of Accounts
• Reduce FX exposure globally with the use of a single set of exchange
rates (over 280 combinations)
• Global regulatory reporting requirements
5
Daily Processing Volumes
• Core Oracle application is the Financial Accounting Hub (FAH)
• The event business model in FAH is used to transform front office trading
system transactions to accounting events
• On the average, we receive 1,500,000 summarized trade transactions daily
from our front office systems
– A single trade transaction is transformed into anywhere from 2-8
accounting transactions
– On a given day, in excess of 20 million records are generated in FAH
across the following tables:
• XLA_TRANSACTION_ENTITIES
• XLA_EVENTS
• XLA_AE_HEADERS
• XLA_AE_LINES
• XLA_DISTRIBUTION_LINKS
• XLA_AE_SEGMENT_VALUES
• GL_IMPORT_REFERENCES
• XLA_AC_BALANCES
• XLA_AE_HEADER_ACS
• XLA_AE_LINE_ACS
6
EBS Technical Performance Metrics
• All trading activity from the prior business day must be processed and
reported on the next business day
• In the U.S. , we are under regulatory requirements to generate
financial statements on a daily basis (i.e., daily closings)
• Under our service level agreements with accounting, all trading activity
must be processed by 5:00 a m. (Chicago time)
• As the earliest we receive files from the trading systems is 2:00 a.m.,
all processing must occur within a 3 hour window between 2:00 and
5:00 a.m.
• To meet these tight SLAs, we choose Oracle Managed Cloud Services
to host our Oracle EBS Applications which run on an Exadata X2-2
7
EBS Technical Architecture
• Hosted by Oracle Managed Cloud Services
• Production Architecture
– Software Platform
• Oracle EBS version 12.1.2
• Database version 11.2.0.4.0 - 64 Bit
• OS – Linux
– Primary Production Environment
• Exadata X2-2
• 2 DB Compute Notes / 7 Storage Cells
• 2 application tiers
– Production DR
– Four non-production environments
8
Understanding our Data Volume
• We anticipated long overnight processing cycles
• In preparation for our first deployment in North America, we
conducted extensive performance tests over four months to right
size our hardware platform
• Before the North America go-live, we anticipated our data growth
to be accelerated and storage requirements to peak in four years.
• To plan for the future, we conducted an assessment of Oracle
EBS archive software vendors with help of an independent
consultant
9
Newedge Storage Projections
• We projected 119 TB of storage used by January 2014.
• This was under the assumption that un-summarized trade
transaction would be brought into Oracle
• Only 18 TB were budgeted
10
Database Table Distribution
XLA_AE_LINE_ACS 12%
XLA_DISTRIBUTION_LINKS 19%
XLA_AE_HEADER_ACS 0%
XLA_AE_LINES 17%
XLA_AE_HEADERS 13%
GL_IMPORT_REFERENCES 10%
XXNE_F10_GL2_COMBINATION 4%
XLA_EVENTS 3%
XLA_TRANSACTION_ENTITIES 2%
Remaining Tables 16%
Database Table Distribution
85% of database size represented data from FAH / XLA tables
11
Impact of our Data Volume
• After three years and when all regions were live, we started to
experience longer batch processing cycle times and a
degradation of performance for reporting
• General system stress and occurrence of software bugs
• Storage costs continued to increase
• The size of our production database grew from 2 to 8 TB. The
data in the FAH application represented 85% of our storage
12
FAH Volume Analysis (2011 – 2014)
Summary by Region
Region Number Records %
United States 14,100,000,000 94%
United Kingom 450,000,000 3%
France 450,000,000 3%
15,000,000,000 100%
13
Data Reduction Roadmap
• Objective
– To reduce storage costs
– To improve system performance
– To provide a foundation for future state archive / lifecycle
management activities
• Major Activities
– Internal projects
– Global data retention requirements gathering
– FAH Archive and Purge
– Database Reorganization
14
Data Reduction Roadmap
– Internal Projects
• Aggregated inbound transactions from trading systems to lower
volume coming into EBS / FAH (versus detail trades)
• Purge FAH interface tables (GL_XLT). These tables consumed a
larger amount of storage than anticipated. New tables were created
with each nightly batch process and were not purged without explicit
maintenance
• Remove inbound and outbound interface files from test
environments
• Apply compression to custom reporting tables.
• Reduced foot print of non-production environment s by sub-setting
after instance was refreshed from production
15
Data Archive Requirements / Strategy
• We needed to understand how much data could be archived
• It was essential to engage the business to define their data retention
requirements
• We asked the business to identify different types of data using the
following data retention categories:
1) Real Time Access. The data is available immediately, directly
accessible from the production system
2) Near Real Time Access. The data is available immediately from an
archive database, but with slower performance and response time than
the production system
3) Offline Access. The data is available for recall from an offline data
storage. Within 48 hours
4) No Data Retention: Data does not need to be retained for this time
frame
16
Database Reorganization
• Data reorganization was necessary to realize the storage reduction from the
FAH Archive / Purge project
• Performed by OMCS
• Reorganization of entire production database
• Objects impacted were tables, segments, and indexes
• The largest schema reorganized were “Apps” and “Bolinf” (custom schema)
• We required reorganization to be completed in under 24 hours
• During re-organization, primary database was disconnected from the stand-
by environment (DR)
• Extensive validation activities were performed
– All indexes were rebuilt and validated
– Reviewed for invalid objects
– Thorough testing to ensure no impact on production processes
• Stand by / DR was rebuilt
17
Actual Storage (2011 – 2014)
• By mid 2013, we realized storage was at an unsustainable level and
needed to be reduced
FAH Purge
Database
Reorganization
18
Storage Savings After Data Reduction
Initiative Production / DR
Storage Savings
Non-Prod Storage
Savings
Total Storage
Savings
1 - BAU Maintenance
0.5 1.0 1.5
2 – FAH Data Archive / Purge
6.4 3.6 10.0
3 - Database Reorganization 2.0 6.0 8.0
19.5
19
FAH Data Purge Project
• Revisited 2010 assessment of Oracle EBS archive software vendors
• Vendors Considered
– Software and professional services
• Solutions offered
– Full licensed model
– Subscription basis
– Build to suit: custom software owned by SG
• Cost Estimates
– Ranged from $145 – $180K
– Excludes second year and beyond subscription cost
• Chose professional services
– Lower cost
– Ownership of solution
– No on-going maintenance cost
20
Financials / Archive Logical Architecture
Oracle Business Intelligence Reporting (OBIEE)
Integrated Data Layer / Union Views
Reporting Tables and
Materialized View
Oracle Financials R12
Financial
Accounting Hub
General Ledger
Modules (Fixed Assets, A/P,
and A/R)
Archive / Reporting Database
Archive - Financial
Accounting Hub
Future State:
Archive - General
Ledger
Future State:
Archive - Modules (Fixed Assets, A/P,
and A/R)
Reporting Tables and
Materialized View
22
Background
• Project start date: Nov 11, 2013
• The right resource, at the right time, at the right price
– FAH Architect, PM/Testing Lead, Developer/DBA
• Newedge brought in Avout to help implement a FAH archiving and
purging solution
– The solution will allow Newedge to significantly reduce the size
of the production environment
– The solution will adhere to Newedge’s global data retention
requirements
– The solution will be repeatable, in other words, can execute the
solution again for the following years
– The solution must be deployed in 36 hour window
– The project must minimize impact on users and Newedge
business
23
Benefits & Scope
• Project Benefits – Storage cost savings
– Important element to Newedge’s archiving and purging strategy
– Reporting performance improvement
– Capable of phasing the implementation by splitting archiving from
purging activities
– No downtime required for the archiving step (performed during off
hours)
• Scope
– Archive and Purge Oracle Financial Accounting Hub application data for
all accounting years 2012 and prior
– Provide solution that addresses all regional data requirements
24
Solution Requirements
• Business Requirements
– Meet regulatory filing requirements
– Business Policy Compliant
– Adhere to Newedge Security Policies
– Archive Reporting Solution
• IT Requirements
– Oracle Referential Integrity Compliant
– Data Integrity
– Configurable Solution
– Easily Transport Data
– Repeatable Process
– Contingency Plans
25
Challenges Considered
• Considered Straight Partition Exchange of Data
– Challenge is that SLA tables are partitioned by
Application ID and not accounting date
• Referential Integrity: Needed to ensure reversing
journals that have accounting events crossing over a
year-end would remain with the non-archive records
• Temp table space limitations. Because the solution
required minimum downtime, the archive step would take
place while the system was still operating
• 36 hour cut-over window
26
Solutions Considered
Drop Storage
• Pros
– End to End processing time is
shorter
• Cons
– Removes indexes and
privileges
– Greater risk that Oracle would
not support if issues found
– Moving data from backup
tables into newly created SLA
tables.
Truncate Storage
• Pros
– No movement of data required for
the 2013 year FAH data.
– Lower risk of support issues
– No issues with index rebuild and
privileges
• Cons
– End to End processing time is
longer
27
Solution Components
• Design calls for archiving to take place prior to purging. For validation
purposes we needed to make sure what was in “Archive” was not in “Base”
and vice versa
• Key business rule: Archive accounting dates had to match purge dates
• Archive Process
1. Count records in archive table and base table
2. Create 2 driver tables based on the date ranges that the users would
like to archive (parallel processing speeds up the process). Create
indexes on driving tables
3. Insert data into 11 archive tables by selecting data directly from
related base table and limiting data based on the respective driving
table
4. After necessary indexes and analysis of archive tables, calculate final
counts in archive table and confirm no integrity violations found
28
Solution Components (continued)
Purge Process
1. Count records in the base table
2. Insert records into purge tables
i. If partitioned source table, select data from FAH partition where
data does not exist in the respective archive table and insert into
temp purge table
ii. If non-partitioned table, select data from base table where the
data does not exist in the respective archive table and insert into
respective purge table
3. Count records : purge + archive = base
i. If no match, print error DO NOT proceed
ii. If counts match, proceed to next steps
4. Truncate base table
i. If partitioned table, truncate FAH partitioned
ii. If non-partitioned, truncate full table
29
Solution Components (continued)
Purge Process - continued
5. Insert into base tables
i. If partitioned table, exchange partition from purge table into the base table
ii. If non-partitioned table, insert all date from the purge table into the base
table
6. Compare new base table counts = purge temp table counts
i. If counts are the same, success
ii. If counts are different, then error
7. Rebuild index partitions/ indexes
i. If partitioned table, rebuild index partition(s) on the table
ii. If non-partitioned table, rebuild indexes
30
Project Approach
• Created flexible schedule to work around business
deadlines and constraints
• Divide and Conquer: separated archive work-stream
from purge work-stream
• Daily status reviews of testing progress
– Involved all regional business representatives in
testing
• Formal signoff of testing results
• Clear communication at all levels
31
Testing
• Partnered with business to develop test plans, included regression
testing of key processes
• Involved global testing resources
• Clearly identified roles and responsibilities
• Divided testing between the archive /purge solutions
• Extensive Testing performed (Unit & UAT)
• Daily Status updates
– Reviewed results and discussed next steps
– Utilized green, yellow, red dashboard – immediately addressed
red issues
• Performance testing – Validated performance of key processes
32
Key Deployment Activities
1. Solution Blueprint into Action
i. Provided Operational Procedures Guide
• Sequential steps with expected timing of activities
ii. Configuration Guide
• Provided step by step configurations of the solution
2. Cutover Plan
i. Worked with team to give complete picture of activities and there
dates to be performed by Avout, Oracle, and Newedge with critical
path items understood
3. Contingency Plan
i. Defined a plan to rollback if we ran into any issues with the purge
process
4. Communications – Emails, Calls, Meetings
33
Lessons Learned
• Testing environment should be EXACTLY like production
– Did not properly account for impact of the DR environment
during production cut-over
• Prior to beginning the testing phase of the project, receive
commitment from business on resources
• Importance of advanced planning and communication with OMCS
34
Questions
• Questions?
• Contact Information:
– George L. Somogyi, Newedge
• E-mail: [email protected]
– Rey Mendez, Avout
• E-mail: [email protected]
35