Appunti per l'esame di Systems and Methods for Big and Unstructured Data - parte 2

PATA

love NESS

LB US

JERS CONTENT)

200

USERS

TRANSACTIONAL ·

ANALITICAL PROCESSING

Mavide andsze

to Du

ARCHITETTURE

DU - single-loge stad

: ance

MENTS dato

devived

real-time

loyer

R/s reasciled

: -

-

E SUPPOR

DECISION (DSS)

SYSTEM

nu istructured

Se olAp

9 bu endais

sonces and ,

Intellation operational

dretadata

siste d extravere reyetive

Di's to ,

2

↳ bota mining

Nota mat

EXtrtYor E

Tra Gor Ispreific analysia)

3 B

SOLE

Sare

Deu Models

(0

Fact

N)

, #ensone t

Decision

s

"

s

↳

1 11 Entity/momized

men

. not

normalized

APORDACHES

SISTEM

LAP

O drill-down

-vari ,

Se si

Mais oregated

te

not

,

HOLAPi combination

- of and Rosa

montp

SNOWFLAYE SF ARCHITETTURE

authentication central

occn

fad

the

approach with scolability

automatic for beleine

to

Nosal Dur ,

infostre

on

- metadata

ottimize panity

-

and cost

pergomances mangen

but

appech mange

TRADITIONAL ~sal

APPROAC ,

Soften vintl nachase

sal interface

Computational computational

resares sares botobese Stage

2 DB instances

has

coed-disn-architecture shoy-motive-architecture

alm individua

(cluster

foed of

querie usa

an sorces

praider indipendent

claud

from

by

allocated sf a continua

there DATA

if

Snowpipe INCETTING

- is a plae

leading doto 2

of piprsing

e te dato

extend stage

tobles ad only E

- , 3

ces command

operations

seledilima

case e cate canfique

stage - o Event

Cloud

pipe of

AP

Rest

virtuali senden

so , of

reflicos virtuali to

corrente

scale sorces

D with com

ands for taben

lange

>

- verg

involved

sequently

in Da and

Orben

Join

,

There

sto

↳ partities table serific

to needs

accoding

, CASSANDRI

ROW COWAN

ORIENTED DATABASE

COLAN STORE

↓ to

las data

radifig

database metwams)

(social elevent

in

↓

rosa red

- l read with

detabes

with lage metiple

fe needs

single rad accesses

and

sains queries Universo -

- - band width

+ utilization

doto

dAp

- pipeline

+

Ron-Length Encesine

COMPRESSION eslity

p

: cock

+

adotabes Dota

son stendine

self + compension

is

eneng

KEYSPACE BASE Disn

will

not seen Time

values -

and

FAMIL

War

- of

-cost inte

alls regetion

ensto and

11

- tugle cost

colum notruction

of the - e

mare of

cost rading

hey-velve ofetoch -

->

newny uniglug RDBM

rach CASSANDRA

is

nau-aihted neur

dentified matelen

by hegwous

i mate-chae

a ,

edim and com

gory te avvildaility

Melication

Salt Apa MISAL structure

/get-slice/ unstructured

CASSANDRA

get Rode

WRITES JoGB RAC dota

a Bi

Exatota

zooms e

4

112

· Integrated

multiget-slice/ stach

soms Quale

1945

sist Biltis

fati in-manag

getangen Cahima i -immag

n Joia

diest incht/batch-mitste No

Ja in

.

emare REPLICAOLACEMENT STRATEGIES

CASSANDRA Warm

WRITE- -

. -

then

talment replica

fact pat-end and

mode

I Simple

- repocaly

Mau

tradlg I

versistent

- datacentes with

+

RING

motation t

Ison-hime , local queries

- seie

Bloom FilterInt

mot indexed

subadum ecotacente

ore

~ positive

felse

toflevel chum

- negative

false

duplicates) never

denomaized (with

- quarz

linean OF OPERATIONS

READ/WRITE

LEVELS

good scolabilitig

elastic

- , , antonelquorum/lo Quah/EAM

CAL

distribution Querin/ALL

egy RIN

- 3 -

consistencig

~ =

BAIE

- N

>

190sig

ortonaic she mapping

-

USE CASES

fost write

durability

dota

Lets of structure

data

Chit

haduar

del considections P dable Bodlan

float

vorint , ,

,

, HAdop

HADOOP-O UDES HAASE

MADREDUCE APACKE ELOSISTE

KADOOP

-Scolable Amarl manager

:

r

Economia

-

Efficient

-

-Relidle i

RDBMS -manage

YARN

Pres anlig

2

MAp -

REDUCE

i

GGA n Kodavn

↳

PETA

interactive MDFS sistem

botch file

botch :

woRnT

RWMT Rebolance that

intequity stard modes

block

lav in

mas

Kieh ore detect

morths6 coord dota

the

regliss failed

star

to (ellow of

linsa i

Nome medesi new pages

saling

line

man data fartbeats

actual faile

Dater e se

mode

structured a

,

Isti

un has deensum

↳ CrCIL

o

MAPREDUCE handline

schokling

dato 40

automatic

fault Lead distribution

fondblization

Folerant balancing (hefte)

platter autat committen

combine

vedue

imput mappare e

o

e spark

SPARK PLETFORM

COMPUTATIONAL erpipeline

United blimps

intective sau

: ques an Grocht

dato

histerial MLib

treamina

proting spensar

intractive : aratica

red time span

foling

stroming (merape

Wh mese

querie

frolingi

sophisticated Tahon

distrial

dolo Hofs

-

OPERATIONS FILE FORMATS

NEW

tanforation

octions -

1 - Parquet

- o

collect wide

How est Mande

sa

, , partitions

in

in patition