System Architecture

MapReduce For Huge Data Precessing. (Cluster.). Key word from X yahoo

http://research.google.com/roundtable/MR.html
http://labs.google.com/papers/mapreduce-osdi04.pdf
http://wiki.apache.org/hadoop/HadoopMapReduce
http://en.wikipedia.org/wiki/MapReduce

Scalable Architecture by Founder and CTO Joyent, Joson Hoffman. Cloud computing….., Direct from Palo Alto,Silicon Valey.

Dr.Jason Hoffman http://joyent.com/
Scalble Architectures

Scaling web Service Applicaction.
   - Scalability, Throughput and Perfomance
   - Limit ( Practical and theoretical )
   - Rules of Ten
   - Web Applications are stateless.

Scalability is Not Performance
Scalability is Not Language Choice

Scalablility is Architectural.
   Scale is Simple.
      - Stack of one lego is one lego in size
      - Stack of te legos is 10x in size

Type of Scalability
  -> Load
  -> Geograhic
  -> Administrative
 
General rules
  - virtualizeation
  - Separating hadware components
  - Keep the hadware setup simple
  - Things should add up
  - Configuration management and distributed control
  - pool and split
  - Understand what each componet can do as maximum 

Fundamental Limites
 - money
 - time
 - people
 - Experience
 - Power
 - Bandwidth

100 Mbps
 - 120 Kb page with 20 objects
 - 2000 req/s constant
 - 8.6 million page views per day

200 kilowatts <= 400 servers

Rules of Ten
 -Tiers
      -> Tiers are different functionally
     -> Tiers should be 10x different in troughput
 - Costs
     - Infrastructure costs=<10% of “Revenue”
 Way to scale
  State = Vertical
  Stateless = Horizontal

Horizontal The Network Traffic Direction.

   -
** Minimize Request
** Signin.domain.com
** Admin.domain.com

Tiered’ed and Silo’ed
  - Silos are DNS-based
 - Static Silo
 - Dynamick Silo
Data Silo ( Files, Objects, Data warehousing )