written by margo seltzer presented by mark simko
TRANSCRIPT
Intro Relational Prehistory New Frontier Emerging Data Needs Flexible Solutions Modularity Configurability Questions
Number and variety of computing devices increasing rapidly
Applications impossible a few years ago are popping up everywhere
Greater need for data storage and retrieval
Examples Gaming applications GPS Messaging applications
IBM and University of California @ Berkeley circa 1970s
Programmers had to rewrite large chunks of code whenever database changed
Solved the problem in two ways Hid the physical organization of DB from
application, only saw logical view Used a declarative language to describe
data, rather then a collection of function calls
1998 DB researches noticed DBMS were becoming too complex
Configuration and management needs to become automated
Become more modular and simple, component-based building blocks
“One size no longer fits all” Michael Stonebraker
Data Warehousing Directory Services Web Search Mobile Device Caching XML Management Stream Processing
Back to basis approach Every application build its own data storage
serviceLots of data management options Each addresses a particular application class,
SQL is used to hide different capabilitiesProduce storage engine that can be configurable Individual can tweak it to their own
requirements Allows concentrated investment in a single
storage system, improving quality
Two properties a solution must possess to address the wide range of emerging applications
Modularity Configurability
Tools allow you to manage
Size Allowing smaller, simpler applications Complexity Some programs are low concurrency,
others medium and some high
Enables the applications and data management capabilities to interact seamlessly
Refers to how well a system can be matched to its environment and application needs
For discussion, this article covers Hardware Environment (operating system) Applications software architecture “Natural” data format of application
Applications software architecture Single thread Collection of cooperating processes (each
single threaded) Multiple threads of control in a single
process Multiple multithreaded processes Event-based architecture
“Natural” data format of application
Physical Clustering Indexing Mechanism Internal Structure of items in the
database
Old style database systems solve old style problems
We need new style database systems to solve new style problems
Conventional Database Management Systems aren’t going away, but we need configurable Database systems to solve today’s and tomorrow’s problems