Single Cores to Big Data - INF 128

basketontarioElectronics - Devices

Nov 2, 2013 (4 years and 2 months ago)

65 views

INF 128: Principles of Informatics

Computation II

From Single Cores to Big Data

INF 128: Principles of Informatics

INF 128: Principles of Informatics

HARDWARE

INF 128: Principles of Informatics

Single storage

for both
data and commands

Storage
separate

from control unit

3

Von Neumann Architecture

http://en.wikipedia.org/wiki/Von_Neumann_architecture

CPU

RAM

von Neumann

The Brain

INF 128: Principles of Informatics


Allows for communication between the
different devices


The wires that connect your hardware
components together


4

The Bus

INF 128: Principles of Informatics


Random Access
M
emory
(RAM)


RAM
contains


code:

currently
running programs


d
ata:

used by those programs.


RAM is
measured in
bytes
.


A byte consists of eight
bits

that may be
either
on

(1) or
off

(0).

5

Memory (Primary Storage)

INF 128: Principles of Informatics

Disk
drives


Hard
Drives


Solid State Drives (SSDs)

Optical
drives


CD, DVD, Blu
-
Ray


Slow, medium capacity

Tape drives


Slow, high capacity


Used for backup

USB Flash drives


Slow, low capacity


Portable

6

Storage
Devices (Secondary)

INF 128: Principles of Informatics

7

Input
and Output Devices

INF 128: Principles of Informatics

CPU

Fetch Next
Instruction

Decode
Instruction

Execute
Instruction

INF 128: Principles of Informatics

Motherboard (Bus)

INF 128: Principles of Informatics

More transistors =


Cheaper CPUs


Higher speeds


More features



More cache

10

Moore’s Law


Number of transistors
doubles every 18 months

INF 128: Principles of Informatics

Improvements in CPU Clock Speed

INF 128: Principles of Informatics

Computation Produces Heat

INF 128: Principles of Informatics

Processor Power Density

4004

8008

8080

8085

8086

286

386

486

Pentium®

Pentium Pro

1

10

100

1000

10000

1970

1980

1990

2000

2010

Year

Power Density (W/cm
2
)

Hot Plate

Nuclear

Reactor

Rocket

Nozzle

Sun’s

Surface

Source: Intel


Nov 19, 2009

13

Pentium II

Pentium III

Pentium IV

INF 128: Principles of Informatics

The Single CPU is History

What can you do with more transistors
at same speed?


Multicore

CPUs


Put multiple processors on a single chip


Run different programs on each core


Some programs can be rewritten to run

on multiple cores:
PhotoShop
, bzip2


Most of you are running multi
-
core


INF 128: Principles of Informatics

End of Moore’s Law

Si ≈ 0.3 nm

INF 128: Principles of Informatics

How much faster can we get??

Parallel Pizzas:


Pizza bakes for 12 minutes


Having 12 pizzas in 12 ovens can’t make any faster


Homemade pizza dough takes 30 minutes


Having more flour or more people can’t speed this up

Speedup is limited by tasks that can’t be divided


“One
-
person” tasks can’t be speeded up

INF 128: Principles of Informatics

Ahmdal’s

Law

𝑆𝑝 𝑢𝑝
=
𝑠
+
𝑝
𝑠
+
𝑝
𝑁

s

is
time taken for one
-
person or sequential tasks

p

is the time taken for the parallel tasks

N

is the number of processors

INF 128: Principles of Informatics

Amdahl’s Law

INF 128: Principles of Informatics

Making a Pizza


You have are having friends over for dinner and you
have agreed to make your wildly popular chicken and
pineapple pizza


Following the directions exactly requires you


Make dough and let rise (30 minutes)


Make sauce (20 minutes)


Cut up chicken strips (10 minutes)


Cut up fresh pineapple (10 minutes)


Shredding fresh mozzarella cheese (10 minutes)


Shred a bit of
habanero

cheese (5 minutes)


Put it all together (10 minutes)


Bake (15 minutes)




Total time: 110 minutes


INF 128: Principles of Informatics

What type of speedup can we get for
making pizza??


What is S?


What is P?


N?

INF 128: Principles of Informatics

CLOUD COMPUTING

INF 128: Principles of Informatics

What is the Cloud?

Cloud computing

is the
on
-
demand use
of
a shared pool
of computing resources
(hardware and software) that
are delivered as a service
over a
network
.

INF 128: Principles of Informatics

Cloud Service Architectures

INF 128: Principles of Informatics

Software as a Service

Clients use provider’s applications running on
cloud infrastructure via the web. Client can
configure application, but not network, OS,
storage, etc.

INF 128: Principles of Informatics

Platform as a Service

Clients can deploy client
-
developed or acquired
applications that were created using
programming tools and languages supported by
provider.

INF 128: Principles of Informatics

Infrastructure as a Service

Client can provision processing, storage,
networks, and other basic computing resources
on which the client can run arbitrary software.

INF 128: Principles of Informatics

IaaS

enabled by Virtualization


Allows easy sharing of hardware resources.


Produces higher utilization rates.

INF 128: Principles of Informatics

IaaS

enabled by Virtualization

Linux

FreeBSD

Windows

Server 2008

Physical

Machine

Virtual

Machines

INF 128: Principles of Informatics

Where is the Cloud?

INF 128: Principles of Informatics

Google DC in The
Dalles
, OR

INF 128: Principles of Informatics

Microsoft Chicago Data Center

INF 128: Principles of Informatics

Rack
-
Mounted Server Closeup


INF 128: Principles of Informatics

Server Internals


INF 128: Principles of Informatics

BIG DATA

INF 128: Principles of Informatics


INF 128: Principles of Informatics

What is Big Data?

Volume:
too large to
store or process on a
single machine.

Velocity:
new
data arrives
quickly.

Variety:
different
types from
organized DBs to
unstructured
docs.

BIG

DATA

INF 128: Principles of Informatics

Big Data: Social Networks


INF 128: Principles of Informatics


INF 128: Principles of Informatics

Big Data: Climate Change


INF 128: Principles of Informatics

Key Points


Moore’s Law
states that transistor count doubles every 18 months (due to
shrinking transistor sizes)


Leads to constantly better cheaper electronics.


But transistors can’t be smaller than an atom…


Heat stopped clock speeds from increasing, so we made
multicore

chips with
multiple CPUs in one package


Speedups are limited by the percentage of the computation that must be
sequentially executed.


Cloud computing
is on
-
demand use of hardware and software located at data
centers around the world.


Software as a Service


Platform as a Service


Infrastructure as a Service


Big data
is defined by 3Vs: Volume, Variety, and Velocity.