Open Source for DBMS? - Microsoft Research

righteousgaggleData Management

Jan 31, 2013 (4 years and 6 months ago)

135 views

Asi
LOWELL
mar++

Michael Franklin

UC Berkeley

© 2003 Michael J. Franklin

The Last DBMS Implementation Project


We need to build one more DBMS


Big 5, 4, 3, … systems not meeting customer
needs.


Attributes


Open Source


Componentized


Cluster (blade server)
-
enabled


Shared Disk


Caching to enable Centralization


Open Source

© 2003 Michael J. Franklin

Why a new DBMS?


Commercial systems not architected for
the emerging world of:
Linux, blade servers, virtualized compute
& storage pools.


Cost of hardware shrinking while cost of
software is not.


Users feeling locked in to one or two
vendors.


Lack of innovation at the DBMS level.


Open Source will enable quick and
continual evolution.

© 2003 Michael J. Franklin

Open Source for DBMS?


Linux is winning the Server Wars.


Driven by CIOs concerned with cost,
reliability, security, and lock
-
in.


It’s just a matter of time until it takes over
the desktop.


Driven by developing nations
-

cost and
innovation on open platform, and hurting
companies (e.g., Sun).


Apache, SendMail, Gnu tools, …


DBMS is doable (but not ready)


MySQL, PostgreSQL, SleepyCat,…


Open Source ensures it will be the last
DBMS needed to be built.

© 2003 Michael J. Franklin

Project 2
-

The Info Sponge


Not a new idea:


Rufus (the original info sponge), Lifestreams, My
Life Bits, Memex…


Personal and Group Information Organizer


Lots of fun Problems


mixture of structured, semi, and un


inherently multimedia


time as a basic organizing principle


place as a basic organizing principle


annotations and their propagation


collaboration


archiving and forgetting


privacy and security