WHAT THE MARKET-LEADING DBMS VENDORS DON

1

• WHAT THE MARKET-LEADING DBMS VENDORS DONT WANT
  YOU TO KNOW
• Disruption is gathering steam

2
Curt Monash

• Analyst since 1981
• Covered DBMS since the pre-relational days
• Also analytics, search, etc.
• Own firm since 1987
• Publicly available research
• Feed at www.monash.com/signup.html
• Blogs, including www.dbms2.com
• White papers and more at www.monash.com

3
Database diversity

• Mike Stonebraker, PhD
• One size doesnt fit all
• Curt Monash, PhD
• Horses for courses
• Database diversity
• Mike and Curt
• The world needs 9 to 11 different kinds of data
  management software

4
(No Transcript)
5
Large enterprise DBMS portfolio

• Principal OLTP/multipurpose DBMS
• Principal OLAP DBMS
• Midrange OLTP/multipurpose DBMS
• Search
• Legacy DBMS
• Other specialty data management

6
(No Transcript)
7
Midrange OTLP/multipurpose DBMS

• Standard editions
• Oracle, DB2, SQLServer, Informix SE
• Deliberately crippled
• VAR-centric
• Progress OpenEdge, Intersystems Cache
• Accidentally crippled
• Open-source
• MySQL, EnterpriseDB

8
OLTP DBMS worries

• Besides the greatest horror data corruption
  concerns include
• License/maintenance cost
• Performance/scalability
• Ease of administration
• Ease of programming
• Reliability/uptime
• Security
•   

9
Three major kinds of transactions

• Traditional business transactions
• Orders
• Employment changes
• Compliance/risk monitoring
• Simple events sensors, logs, etc.
• Web site clicks
• Network events
• Device monitoring
• Vehicle monitoring
• RFID
• Content serving

10
Traditional business transactions are

• Complex
• Consistent in the face of complexity
• Stringently industrial-strength
• Real business need
• Customer expectations
• Compliance

11
Issues to consider for applications that record
complex transactions

• Schema complexity (integrity)
• Schema variability
• Peak performance
• Uptime
• Security

12
Issues to consider for applications that record
simple events

• Performance
• Uptime
• What happens to the data next?

13
Issues to consider for applications that serve
content

• Which datatypes?
• Scale
• The alphanumeric parts

14
Application metrics

• Peak concurrent update throughput
• Query complexity and volume
• Transaction (and constraint!) complexity
• Overall database size (and change!)
• Uptime requirements
• Security/compliance requirements
• Datatype needs

15
And how will those evolve?

• Business model changes ?
• Functional changes

16
Environmental considerations

• Hardware (SMP, blade, toy collection)
• Middle tier
• DBMS expertise (and where it sits in the
  organization)
• Database administration tools
• Development tools
• Fixed-point applications (and how good is their
  generic JDBC/ODBC support?)

17
And how will THOSE evolve?

• Consolidation -- but what does that mean in your
  shop?
• Modularity

18
Example 1 Compliance/risk monitoring

• Many feeder systems
• One schema per feeder system
• Accept both relational ETL and XML
• Output via BI

19
Key requirements 1

• Rigorous security
• Easy administration
• Eventual XML support
• Unknown scalability

20
Example 2 Contractually-defined products

• Complex financial instruments
• Vacations
• Warranties

21
Key requirements 2

• Strong native XML
• Complex constraints
• Availability
• Security
• Volume?

22
Example 3 Content sharing and selling

• Web-facing video, music, photo, etc.
• Internal content management

23
Key requirements 3

• Performant media datatype support
• Performant order entry
• Performant user tracking and personalization
• Spike scalability
• 24/7 availability

24
Major areas of OLTP DBMS differentiation

• Performance and scaling
• Administration and 24/7 operation
• Constraints and referential integrity
• Triggers and stored procedures
• Datatype support

25
Performance and scaling

• Baseline, peak, future
• For which features?
• How sub-linear?

26
Administration and uptime

• Ongoing functions backup, security, etc.
• Indexes and mandatory maintenance??
• Replication, fail-over, etc.

27
Database constraints

• What can be done in theory?
• Does it perform?

28
Triggers and stored procedures

• Performance
• Languages
• Automatic generation
• Development, debugging, maintenance

29
Datatype support

• What do you need?
• Performance
• Datatype extensibility
• (Where relevant) Quality of search

30
Todays main topics

• You can and should use multiple DBMS
• In particular, midrange OLTP DBMS are appealing
• Not all midrange OLTP DBMS are created equal
• Both application and environmental considerations
  are important
• More info at www.monash.com and www.dbms2.com