E conomics 101 Demand and Supply Equilibrium (think load balancing) - - PDF document

e conomics 101
SMART_READER_LITE
LIVE PREVIEW

E conomics 101 Demand and Supply Equilibrium (think load balancing) - - PDF document

O utline Introduction Food for thought Mariposa: a wide-area distributed Economics 101 database system Paper - background Previous assumptions Key issues with assumptions New requirements M. Stonebraker et al.


slide-1
SLIDE 1

1

Mariposa: a wide-area distributed database system

Presentation: Nikhil Discussion: Albert

  • M. Stonebraker et al.

UC Berkeley

Outline

Introduction Economics 101 Paper - background

  • Previous assumptions
  • Key issues with assumptions
  • New requirements

Paper - approach

  • Mariposa – model and architecture
  • Business process
  • All other stuff (storage, naming…)
  • Results

Epilogue

Food for thought

Introduction

What is D in DDBMS?

Requirements Why do we care? Some examples (90’s and beyond)

Economics 101

Demand and Supply Equilibrium (think load balancing)

http://www.sci.wsu.edu/math/Lessons/SupplyAndDemand/theory.html

Outline

Introduction Economics 101 Paper - background

  • Previous assumptions
  • Key issues with assumptions
  • New requirements

Paper - approach

  • Mariposa – model and architecture
  • Business process
  • All other stuff (storage, naming…)
  • Results

Epilogue

Food for thought

Previous (wrong!) Assumptions

Static data allocation Single administrative structure Uniformity

slide-2
SLIDE 2

2

Key Issues (with assumptions)

Static data allocation

Is it distributed?

Single administrative structure

No free will

Uniformity

Surreal

New Requirements

For non-uniform, multi-admin WAN

Scalability Data mobility No global synchronization Total local autonomy Easily configurable policies

Food for Thought

Starter

What applications can you imagine using

Mariposa for?

What about systems using the previous

assumptions for distributed system?

Outline

Introduction Economics 101 Paper - background

  • Previous assumptions
  • Key issues with assumptions
  • New requirements

Paper - approach

  • Mariposa – model and architecture
  • Business process
  • All other stuff (storage, naming…)
  • Results

Epilogue

Food for thought

Account Budget Broker Bid Bidder Rush

  • An Introduction to the Rush Language (A. Sah, J. Blow, and B.

Dennis)

  • Unlike ‘if’ -> tests a predicate at one point
  • Like an ‘if’ which is always watching
  • n { $x < 10 } do { puts stdout $x }

Stride (handicapped ||)

Mariposa – Terminology

Mariposa – Economic Model

ABC of market economics

Budget allocation Advertising service Join/leave system (fragment trading) Buy/sell => bidding

slide-3
SLIDE 3

3

Mariposa – Economic Model

Client 1 Client 2 Client 3 . . . . Client n Site 1 Site 2 Site 3 Site 4 Site n . . . . Broker Retail Market

  • Insti. Market

Exit Enter

Mariposa – Architecture

Client Middleware Local EE

Mariposa – Architecture

Client Middleware Local EE

Mariposa – Architecture

Client Middleware Local EE

Mariposa – Architecture

Client Middleware Local EE

Mariposa – Business Proc.

Free market at its best

Players (clients) willing to buy data at a

certain price

Players (bidders) willing to sell data at a

certain price

Neutral mediator (broker) to enforce

fair-play

Broker

Expensive bid protocol Purchase order protocol

slide-4
SLIDE 4

4

Mariposa – Business Proc.

Bidding

Bid acceptance Search problem Coalesce (in strides) and then evaluate Heuristics (at last!) Finding bidders Good advertisement leads to higher sales ☺ Setting bid price (auto load balancing) Network bidder

Food for Thought

Main Course

Does this bidding process seem like a

good model to you? Can you think of

  • ther applications for which it might be

relevant?

Mariposa – Storage Mgmt

Idea => Maximize revenue

Fragment trading Splitting and coalescing

Names

Internal Full Common Name contexts

Name resolution Name discovery

Mariposa – Naming

Tests and Issues

Environment

3 relations, 3 sites, 11 MB (!) of data Purchase Order Vs Expensive bid Bidding

Issues

Market eco ^ AI => only simplistic

possibilities

Outline

Introduction Economics 101 Paper - background

  • Previous assumptions
  • Key issues with assumptions
  • New requirements

Paper - approach

  • Mariposa – model and architecture
  • Business process
  • All other stuff (storage, naming…)
  • Results

Epilogue

Food for thought

slide-5
SLIDE 5

5

Epilogue

Mariposa – where is it now?

  • Mariposa > Cohera > PeopleSoft > Oracle (But!)

DDBMS – commercial and deployed

  • None that I know of

DDBMS – current research

  • P2P systems
  • Semi-structured and Un-structured data
  • What Can Databases Do for Peer-to-Peer?
  • The state of the art in distributed query processing

Food for Thought

Dessert

How does Mariposa compare to

today's P2P's system? How is it the same? How is it different?

Thank You