2. DBS & MariaDB
Corporate Banking &
Future Ready Technology
26 February 2018
2
The presentations contain future-oriented statements, including statements regarding the
Group’s vision and growth strategy in the light of anticipated trends as well as economic
and market conditions. Such statements necessarily involve risks and uncertainties,
which may cause actual performance in future periods to differ from projections.
3. Disclaimer: The information contained in this document is intended only for use during the presentation and should not be disseminated or distributed to parties outside the presentation. DBS
Bank accepts no liability whatsoever with respect to the use of this document or its contents.
Be Digital to the Core
7. Systematic multi-year effort to streamline
tech infrastructure
Customer focus and culture transformation
internally
Accelerated income growth driven by scale,
speed & efficiency
Gandalf Tech Transformation
Culture Transformation
Tech-driven business
1
2
3
Gandalf Technology and Culture Transformation
Moving from legacy company to tech company
9. Teller
Internet & Mobile Banking
Corporate Internet Banking
Customer Relationship Mgt
Customer
Private banking
Liquidity Management
Remittances
Unsecured loans
Secured loans
Trade Finance
Factoring
Treasury
Mgt Information Sys
Credit Approval
Finance
Singapore Hong Kong Taiwan Indonesia ChinaIndia
Overseas
BranchesCHANNELSPRODUCTS&SERVICESENABLERS
DBS application landscape in 2009
10. Teller
Internet & Mobile Banking
Corporate Internet Banking
Customer Relationship Mgt
Customer
Private banking
Liquidity Management
Remittances
Unsecured loans
Secured loans
Trade Finance
Factoring
Treasury
Mgt Information Sys
Credit Approval
Finance
Singapore Hong Kong Taiwan Indonesia ChinaIndia
Overseas
BranchesCHANNELSPRODUCTS&SERVICESENABLERS
By 2017…We’ve migrated to MariaDB
11. Migrated CAT 1 / 2 type applications66
Migrated CAT 3 / 4 type applications82
ELA signed with MariaDB to strengthen partnership2+1
Features introduced to make migration seamless30
Developers who are trained in MariaDB500
Our 2nd Anniversary with MariaDB
12. 0
100
200
300
400
500
600
700
800
Q4 2016 Q1 2017 Q2 2017 Q3 2017 Q4 2017 Q1 2018
INSTANCES
Key Banking Domains
using MariaDB
Payments
Trade
Cash
Channels
Digital Bank Initiatives
Customer
4
X
700+ total deployments
54% critical applications
30+ Oracle features enabled in MariaDB
Increase
MariaDB Deployment Footprint
13. Once schema change is completed on cluster2, it’ll
be merged with data/records change in cluster 1.
Cluster 2 will be turned over to production.
(A) + (B) (A’) + (B)
(B’)
(A’)
(A’) + (B’)
Cluster 1
(blue)
Cluster 2
(green)
1
Transactions continue to be supported
by cluster 1, while schema changes
are applied to cluster 2.
Cluster 1 is converted to
standby mode.
3
2
Currently, no zero downtime deployment for database schema change
Ideally, continuous transactions while schemas changed using standby resources
Then, to merge the schema change once verified as successful
Blue/Green Deployment for MariaDB
14. 14
Design template
Config parameters
Build profile Registry
Shared DB
Users
DevOps
Blue
Green
Switch
Note: B-G for
DB still under
exploration.
6
8
432
5
7
1
Container
orchestration
Registry
Blue/Green Deployment: Microservices
config.sit-V1
config.sit-V2
config.sit-Vn
JSONs
App Binary
Certificates
Build Profile
Transformation
Engine
V1
V2
Vn
15. Written Scanned Built
Tested and
verified
Deployed in
different
environments
Manual Automate Manual Automate Manual Automate
Manual Automate
Manual Automate
DEVELOPMENT
ENVIRONMENT
PRODUCTION
ENVIRONMENT
Manual
Automate
Manual
Automate
To introduce changes to production, code needs to
be…
16. 2014 2017
Volume of automated test executions
7.5x
10X
10x
100%
DevOps: % of automated application releases
2014 2017
100%
2014 2017
Cadence – No. of monthly deployments
7.5X
0
50
100
Aggressive automation has enabled us to move faster &
increase efficiency
17. Sequences
Views
Packages
Support for 38 decimal digits
SQL Mode Oracle
Support for PL/SQL
Connection Pool – driver
Solution for Custom Types
Function based Index: Virtual
columns
Solution for PIPE row/
ConnectBy
Support for multiple triggers
Key Enablers
18. 2018 DBS-MariaDB Roadmap
MariaDB asof to
query historical
data
Data analytics
using MariaDB
ColumnStore
Sharding with
MariaDB Spider
Partition
by Status
Enable data
type
compatibility
(ColumnStore)
More
compatibility
certification with
software vendors
(SWIFT, IBM TX;
Kasisto)
Enable
parallelism
for stored
procedure
New Build
Available Features
19.
20. ⬆10% in Customer satisfaction
Nielsen SME CE Survey 2017: 75%, outperforming others’ 56%
2x Faster user experience
Performance improvements
Zero downtime
Capability for maintenance and recovery
3x Capacity volume
Increase in system capacity for growth in customer base
$4.1m Net savings
5 year savings after initial investment
Corporate Internet Banking Platform - Performance
Improvements
22. Allow longer (12 months) online transaction history retrieval
Embrace cloud technology
Minimal downtime
Total $4m net savings over 5 years
Enable Scalability running on virtual machine
Transaction Data Store - Performance Improvements
Gandalf Tech Transformation
Systematic multi-year effort to streamline tech infrastructure has enabled us to rapidly adopt digital technologies and build sustainable competitive advantage
Culture Transformation
Digital technology combined with customer focus and culture transformation improve market share through superior customer experience and create new markets through ecosystems
Tech-driven business
Accelerated income growth achieved with low marginal costs as a result of scale, speed and efficiency
(verbally articulate constraints)
Schema change is a reduction from existing (columns removed or data type changed)
Key violation occurs during the change (we must avoid data change to the cluster 2 (green) until it is turn primary)
Index change
If there is data change in cluster 2 (green), then, there will be no path back to cluster 1 without affecting data integrity
*Application needs to be designed with such limitations in mind to prevent data synchronization breaking during zero downtime change.
LEGEND
Database records change: UPDATE tableA SET column1=value1, column2=value2 where primaryKey=keyvalue;
Database schema change: CREATE TABLE tableA (column1 int, column2 varchar(8), …);
Deployment parameters are managed in separate configuration files.
Deployment Template (JSON) is created using configuration files
Deployment Template, APP Binary certs are uploaded to Nexus
RA builds the image and pushes to the OpenShift registry
RA spawns the container using the images in Openshift registry
Similarly RA can spawn containers with different app versions
IBG Support can determine which version of the application to expose to users
Users access the application via HAProxy
There are over 30 features we work with for MariaDB. These are the key highlights.
Notes:
Theoretical capacity increase on new Linux/Jboss/MariaDB is approximately 3 times.
Old AIX – 4x App Servers (16 JVMs) @ 1,600 CCU
New Linux – 12x App Servers (24 JVMs) @ ~4,500 CCU