• No results found

Open source platform as a service

N/A
N/A
Protected

Academic year: 2021

Share "Open source platform as a service"

Copied!
23
0
0

Loading.... (view fulltext now)

Full text

(1)

Open

source

 

platform

as

a

service

Dana Petcu

Dana

 

Petcu

West

 

University

 

of

 

Timisoara

&

 

Institute

 

e

Austria

 

Timisoara

1

Romania

(2)

Overview

Overview

1. The

 

problem:

 

vendor

lock

in

 

in

 

Clouds

2. The

 

solution?

 

Interoperability

 

and

 

portability

 

3. The

 

role

 

of

 

open

source

 

platforms

4. mOSAIC case

 

studyy

5. Research

 

challenges

6. Further

 

steps

p

2

2

(3)

Vendor lock-in due to proprietary APIs

Vendor lock in, due to proprietary APIs

API

spec

spec

How

 

to

 

protect

 

my

 

application

from lock in?

API

spec

01011

001

from

 

lock

in?

API

spec

3

3

(4)

Interoperability and portability: challenges

Interoperability and portability: challenges

Scenarios

 

using

 

multiple

 

Clouds

Multi

dimensional

 

problem

Federation

01 01 01 01 01 01

POLICY:

Federate,

communicate

of Clouds

011 01 1 01 1

Cloud

RUNTIME:

between providers

01 01 1

InterOp

DESIGN:

RUNTIME:

Migration

support

4

DESIGN:

Abstract the programmatic

differences

Markets of Clouds

4

01 01 1 01 01 1

(5)

Interoperability and portability: approaches

Interoperability and portability: approaches

E.g.

Levels

Techs

E g

S

i

Business

Strategies,

 

regulations,

 

mode

 

of

 

use

Function

 

calls

 

and

 

Domain

 

specific

 

lang.

E.g.

Automated

 

translation

 

in

 

code

Appl &

 

service

Semantic

responses

Automation,

 

configuration

Standards

 

in

 

deployment

 

& migration

Abstraction

 

layers

Semantic

 

repositories

UCI

Mediators,

 

frame

works

 

(SLA@SOI)

Techs

 

&

 

infrastr

Management

&

 

migration

Protocols

 

for

 

requests/responses

Pre

deployment,

 

Open

 

protocols

Standards

OVF/DMTF,

CDMI/SNIA

 

OCCI,

 

Deltacloud

5

Network

Image

 

&

 

data

work

loads

Allocation,

admission

Open

 

APIs

jClouds,

OpenStack

 

libcloud,

 

Open

5

(6)

PaaS

 

– the

 

worst

 

case!

™

each PaaS provider offers a special flavor in its design

™

each

 

PaaS provider

 

offers

 

a

 

special

 

flavor

 

in

 

its

 

design

 

™

not

 

all

 

the

 

features

 

that

 

are

 

expected

 

ƒ

rarely debug facilities

ƒ

rarely

 

debug

 

facilities,

 

ƒ

no

 

Security

as

a

Service

ƒ

often

 

not

 

for

 

private

p

 

clouds

™

portability

 

is

 

possible

 

only

 

between

 

a

 

small

 

no.PaaSs

ƒ

in

 

case

 

of

 

open

source

 

clones

 

of

 

the

 

proprietary

 

ones

 

™

the

 

problem

 

escalates

 

with

 

the

 

increase

 

of

 

no.PaaSs

ƒ

increased

 

no.

 

in

 

last

 

two

 

years

 

6

6

(7)

Deployment

 

of

 

Cloud

 

techs

 

at

 

PaaS

 

level:

 

two

 

types

Platform Service |

Platform Software |

Platform

 

Service

 

|

 

Hosting

 

|

 

Integrated solution

Platform

 

Software

 

|

 

Software

 

service

 

|

 

Deploy

based solution

Integrated

 

solution

™

Well

 

knows

 

examples:

ƒ

Google’s

 

AppEngine,

 

Microsoft’s

 

Azure,

 

R kS

’ Cl

dSi

A

’ B

lk

Deploy based

 

solution

™

Deployment

 

of

 

middleware

in

 

data

 

RackSpace’s CloudSites,

 

Amazon’s

 

Beanstalk,

 

Salesforce’s Force.com,

 

Joyent’

 

SmartPlatform

™

Open

source

 

only

 

to

 

develop

 

appls

centers

 

™

Easy

 

way

 

to

 

deal

 

with

 

portability and

ƒ

to

 

allow

 

customization

™

Appl support

 

– different

 

approaches:

ƒ

Deploy code to specific VMs (Azure Beanstalk)

portability

 

and

 

interoperability

 

(framework

 

category)

™

O

h

th

ƒ

Deploy

 

code

 

to

 

specific

 

VMs

 

(Azure,

 

Beanstalk)

ƒ

Devolop using

 

rules,

 

platform

 

deal

 

with

 

deployment

 

(AppEngine,

 

Heroku)

C

t

t d t t b i t

t d b P S t

7

™

Open

source

have

 

the

 

potential

 

to

 

impact

 

the

 

market

 

as…

ƒ

Create

 

metadata

 

to

 

be

 

interpreted

 

by

 

PaaS at

 

run

time

 

(Force.com,

 

OrangeScapes)

 

7

ƒ

PVM/Parallel

(8)

Open

source

 

Platform

 

Software

Product AppScale Cloud

Foundry ConPaaS mOSAIC OpenShift TyphoonAE WaveMaker

Owner Univ. Ca‐

lifornia VMWare ContrailConsortia mOSAICConsortia RedHat TobiasRodäbel    VMWare

Site appscale.cs. ucsb.edu www.cloud  foundry com www. conpaas. eu www. mosaic‐ cloud eu

open shift.com code. google. com/p/typho onae

www.wave 

maker.com

foundry.com eu cloud.eu onae

Repository appscale. googlecode. com/svn/ github. com/ cloud foundry www.  conpaas. eu/  download/ bitbucket.  org/ mosaic github.  com/ openshift code. google. com/p/ typho‐onae/ downloads/ dev.wavemak er. com/wiki/  bin/

State 1.5/Jul 2011 0.x , Beta 0.1/Sep 2011 0.5/Jun’12, 

Beta Production 0.2/Dec2010/beta/   6.4.4/Dec2011  

Languages Python, Java, 

Go Java,Ruby,Node.js,    Groovy PHP Java, Python, Erlang, Node.js Java,  Python, 

Perl, PHP, Ruby Python Java

Data

Support HBase,Hypertable, Redis MySQL Cl t MongoDB, SQLFire, PotsgreSQL, R di Scalaris, MySQL, XtreemFS Riak,  CouchDB, MemcachDB R di MySQL, MongoDB, Amazon RDS MongoDB, MySQL, Berkeley DB JE Amazon S3, Rackspace Cluster, Cassandra,  Voldermort, MongoDB, Memcache‐ DB Redis Redis,  MySQL, Amazon S3, HDFS JE

OS Ubuntu VMWare XtreemOS Linux Red Hat Debian VMWare

8

OS Ubuntu, 

CentOS VMWareimage XtreemOSimage Linux RedVirtualization Hat  Debian,Ubuntu  VMWareimage

Messaging Channel RabbitMQ Own 

design RabbitMQ Owndesign  RabbitMQ,ejabberd, 

Channel

Own design

Clouds tested Amazon EC2,

E l t VMWare Own testbed AmazonFl i l EC2, RightScaleR k Google EC2,R k

8

Eucalyptus Flexiscale,  OpenNebula. Eucalyptus … Rackspace, Smart‐Cloud, Amazon Rackspace, OpSource, Eucalyptus

(9)

Open

source

 

Platform

 

Software

d

l

d

d

hif

Product

CloudFoundry

mOSAIC

OpenShift

Development support 1 2 3

Dedicated to web aps or general Web apps General Web apps

Desktop Cloud Simulator Yes Yes No

API access No Yes No

Support standard programming libs Yes Yes Yes

Impact on web application architecture No Yes No

Impact on web application architecture No Yes No

Complexity of porting web application  Medium Low Low

Standard support tools Spring Tools No JBoss,  Zend 

Thread access Yes No Yes

MySQL Yes Yes Yes

Allows to choose stack components Yes Yes No

Allow to pullp  data out Yes Yes Yes

Debugging mode Yes Yes Yes

Deployment support 1 2 3

Lock‐in when building own Cloud Yes (VMWare)  No  Yes (RHE)

Web server (e.g. Tomcat) Yes Yes Yes

Build‐in‐balancer No Yes Yes

Auto‐scaling app server No Yes Yes

A t li d t b N Y N

Auto‐scaling database No Yes No

Performance analytics Yes No Yes

Support multiple Cloud providers Yes Yes Yes

Agreements SLA No Yes No

Deploy with a special tool Yes No No

Support Private Cloud Yes Yes No

Allows to add third party components Yes Yes Yes

9

Allows to add third party components Yes Yes Yes

Execution support 1 2 3

Command line (CLI) Yes Yes Yes

Web console No Yes Yes

Access to logs via web No Yes Yes

Web based monitoring No Yes Yes

Multitenant  Yes Yes Yes

(10)

mOSAIC’s

 

marketing

 

motto:

 

“ l i

h

h h

l

d ”

“Flying

 

through

 

the

 

Clouds”

API

spec

Application

Portability!

0101

1001

API

spec

Portability!

0101

1001

API

API

spec

(11)

mOSAIC

 

as

 

R&D

  

collaboration

 

effort

Consortium:

 

www.mosaic

cloud.eu

1.

Second

 

University

 

of

 

Naples,

 

Italy

2.

Institute

 

e

Austria

 

Timisoara,

 

Romania

3.

European

 

Space

 

Agency,

 

France

4.

Terradue SRL,

 

Italy

5.

AITIA

 

International

 

Informatics,

 

Hungary

6.

Tecnalia,

 

Spain

7.

Xlab,

 

Slovenia

8.

University

 

of

 

Ljubljana,

 

Slovenia

9.

Brno

 

University

 

of

 

Technology,

 

Czech

 

Republic

September 2011:    1

st

API implement. (Java)

September 2012:    1

st

stable PaaS,

2

nd

API i

l (P th

)

2

nd

API impl. (Python)

(12)

Layered architecture

Open-source and deployable PaaS

12

(13)

How

 

to

 

use

 

it?

    

One

  

scenario:

™

Write

 

component

based

 

application

 

ƒ

Languages:

 

Java,

 

Python,

 

Erlang,

 

Node.js

 

C

i ti

th

h

i

ƒ

Communications

 

through

 

message

 

passing

ƒ

Respect

 

the

 

event

driven

 

style

 

of

 

programming

™

Debug

g y

 

your

 

application

pp

 

on

 

the

 

desktop

p

 

or

 

on

premise

p

 

server(s)

ƒ

Within

 

Eclipse

ƒ

Use

 

Personal

 

Testbed Cluster

 

using

 

VirtualBox for

 

the

 

VMs

™

D

l

li ti

i

Cl

d

™

Deploy

 

your

 

application

 

in

 

a

 

Cloud

ƒ

Assisted

 

by

 

Cloud

 

Agency

 

and

 

Broker

 

(with

 

SLAs)

™

Monitor

 

&

 

modify

y

 

the

 

applications

pp

ƒ

Control

 

the

 

life

cycle

 

of

 

the

 

components

 

(start/stop/replace)

Need

 

help?

13

Follow

 

documentation

 

and

 

YouTube

 

demos

 

(search

 

“mOSAIC Cloud

 

computing”)

(14)

Application lifecycle

5) Monitoring

Cloudlets + mOS software modules

1) 3) Deployable Apps Application descriptor 2) desc pto 4) Deployment descriptor

(15)

How is designed the vendor independent API?

(J) Component

(J) Component

(P) Component

(P) Component

Classical

 

components

 

of

 

applications

(J) Connector API

(P) Connector API

(J) Cloudlet API

(P) Cloudlet API

Operations

 

with

 

standard

 

Component

 

reacting

 

to

 

events

pp

Interoperability API

Proxies

 

generator

Operations

 

with

 

standard

 

type

 

of

 

resources

Driver

API

API

API

API

Driver

API

 

for

 

same

 

type

 

of

 

resource

Driver

API

API

15

15

D. Petcu, G. Macariu, S. Panica, C. Craciun, Portable Cloud Applications - from Theory to

(16)

Research

 

issues

 

related

 

to

 

mOSAIC’s

 

PaaS

™

Auto

scaler

ƒ

N.M.

 

Calcavecchia,

 

B.A.Caprarescu,

 

E.

 

Di

 

Nitto,

 

D.

 

J.

 

Dubois,

 

D.

 

Petcu,

 

DEPAS

A D

t li d P b bili ti Al

ith

f

A t S li

DEPAS:

 

A Decentralized Probabilistic Algorithm for Auto‐Scaling

,

 

Computing,

 

Springer,

 

doi:

 

10.1007/s00607

012

0198

8,

 

June

 

2012,

 

available

 

at

 

http://arxiv.org/abs/1202.2509

™

Scheduler

ƒ

M.

 

Frincu,

 

Scheduling Highly Available Applications on Cloud 

Environments

,

,

 

Future

 

Generation

 

Computer

p

 

Systems,

y

,

 

May

y

 

2012,

,

 

doi:10.1016/j.future.2012.05.017

™

Naming

 

service

ƒ

S P

i

Di t ib t d R

Id tifi ti

S

i

f

Cl

d

ƒ

S.

 

Panica,

 

Distributed Resource Identification Service for Cloud 

Environments

,

 

submitted

16

16

(17)

Auto

scaler

™

Problem:

 

Most

 

existing

 

auto

scaling

 

solutions

 

are

 

centralized

™

Solution:

™

Solution:

ƒ

based

 

on

 

a

 

P2P

 

architecture

ƒ

one

 

autonomic

 

service

 

is

 

deployed

 

on

 

each

 

VM

™

DEPAS

 

algorithm:

ƒ

each VM probabilistically decides to

each

 

VM

 

probabilistically

 

decides

 

to

 

add

 

new

 

nodes

 

or

 

remove

 

itself.

ƒ

probability

 

is

 

computed

 

based

 

on

 

an estimation of the average

an

 

estimation

 

of

 

the

 

average

 

system

 

load

ƒ

the

 

average

 

system

 

load

 

is

 

approximated by each node with

17

approximated

 

by

 

each

 

node

 

with

 

the

 

average

 

load

 

of

 

itself

 

and

 

its

 

neighbors

(18)

Auto

scaler

™

On

 

Amazon

 

EC2

 

in

 

COMPUTING

 

j.

™

Test

 

conclusions:

ƒ

After

 

a

 

period

 

of

 

adaption,

 

DEPAS

 

allocates

 

a

 

right

 

capacity

 

(between

 

the

 

optimum

 

capacity

 

at

High capacity nodes are

 

added

Low capacity nodes are added

max

 

load

 

and

 

optimum

 

capacity

 

at

 

min

 

load)

ƒ

The

 

delay

 

in

 

adaptation

 

is

 

cased

 

by

 

the

 

relatively

 

high

 

duration

 

of

 

load

 

monitoring

 

timeframe

 

and

 

cycle duration

cycle

 

duration

ƒ

The

 

benefit

 

comes

 

in

 

the

 

high

 

stability

 

(no

 

oscillations)

 

which

 

is

 

impressing

 

for

 

a

 

decentralized algorithm

decentralized

 

algorithm

18

18

(19)

Scheduler

™

Problem:

ƒ

component

based

 

appls can

 

encounter

 

failures

 

of

 

components

ƒ

a scaled application can span its components on several nodes

a

 

scaled

 

application

 

can

 

span

 

its

 

components

 

on

 

several

 

nodes

ƒ

finding

 

the

 

optimal

 

no.component types

 

needed

 

on

 

nodes

 

so

 

that

 

every

 

type

 

is

 

present

 

on

 

every

 

allocated

 

node

ƒ

cost restrictions and threshold for no nodes

ƒ

cost

 

restrictions

 

and

 

threshold

 

for

 

no.

 

nodes

™

Application

 

to:

ƒ

Highly

 

available

 

Web

 

2.0

 

applications

™

Novelty:

ƒ

Schedule

 

components

 

instead

 

VMs

™

™

Scheduling

 

algorithms:

ƒ

One

 

that

 

produces

 

an

 

optimal

 

solution

 

in

 

case

 

when

 

the

 

load

 

of

 

every

 

component

 

is

 

known

19

ƒ

One

 

that

 

produces

 

a

 

sub

optimal

 

solution

 

and

 

relies

 

on

 

a

 

GA

 

to

 

allocate

 

components

 

in

 

case

 

the

 

component

 

load

 

is

 

unknown

(20)

Scheduler

™

Test

 

goals:

ƒ

Ability

 

to

 

achieve

 

high

 

availability

 

measured

 

through

 

reliability

 

indicator

ƒ

Heterogeneity of the load on every node; expect load close to max of

Heterogeneity

 

of

 

the

 

load

 

on

 

every

 

node;

 

expect

 

load

 

close

 

to

 

max

 

of

 

resource

 

capacity

 

20

20

(21)

Naming

 

service

™

Problem:

ƒ

Usually,

 

common

 

naming

 

services

 

are

 

implemented

 

in

 

a

  

centralized

 

manner

ƒ

DNS

 

service

 

uses

 

a

 

static

 

configuration

 

scheme

 

and

 

it

 

is

 

therefore

 

not

 

suitable

 

to

 

be

 

deployed

 

as

 

a

 

PaaS component

™

Solution:

ƒ

distributed

 

naming

 

and

 

resource

 

identification

 

service

 

ƒ

allows

 

a

 

client

 

to

 

consume

 

or

 

communicate

 

with

 

a

 

resource

 

without

 

having

 

to

 

deal

 

with

 

resource

 

location

 

or

 

discovery protocols

discovery

 

protocols

 

ƒ

create

 

a

 

distributed

 

system

 

on

 

top

 

of

 

the

 

DNS

 

service

 

that

 

manages

 

the

 

naming

 

ops

p ( g

 

(register,

, p

 

update

 

and

 

retrieve)

)

 

in

 

a

 

21

distributed

 

manner

 

and

 

by

 

using

 

the

 

(22)

What’s

 

next?

™

Full

 

integration

 

&

 

testing

 

&

 

benchmarking

 

&promotion

of mOSAIC PaaS

deadline Spring 2013

of

 

mOSAIC PaaS deadline

 

Spring

 

2013

™

Develop

 

further

 

the

 

open

source

 

and

 

deployable

 

PaaS:

ƒ

[RO]

 

Improve

 

platform

 

services

  ‐

automated

   

management

 ‐

in

 

the

 

//

frame

 

of

  

AMICAS

 

(http://amicas.hpc.uvt.ro),

  

2012

2014

ƒ

[EC

FP7]

 

Develop

 

HPC

 

services

 

in

 

Cloud

 

using

 

parts

 

of

 

the

 

mOSAIC’s PaaS

in

 

the

 

frame

 

of

 

HOST

 

(http://host.hpc.uvt.ro),

 

2012

2014

™

Develop

 

top

level

 

services

 

based

 

on

 

the

 

PaaS:

ƒ

[EC

FP7]

 

MODAClouds

model

driven

 

engineering

 

in

 

multiple

 

Clouds

 

(http://www modaclouds eu) 2012 2015

(http://www.modaclouds.eu),

 

2012

2015

  

™

Develop

 

applications

 

that

 

are

 

using

 

the

 

PaaS:

ƒ

[EC

CIP]

 

SEED

  ‐

e

Government

 

services

 

on

 

Clouds,

 

2012

2014

22

(23)

Other

 

references

D P t H t b ild li bl OSAIC f lti l l d i EWDCC '12 2012 •D. Petcu, How to build a reliable mOSAIC of multiple cloud services, EWDCC '12, 2012

•D. Petcu et al, Towards Component-based Software Engineering of Cloud applications, WICSA 2012 80-81, 2012. •B. A. Caprarescu et al. Decentralized Probabilistic Auto-Scaling for Heterogeneous Systems, ADAPTIVE 2012, 7-12 •D.Petcu, J.L. Vazquez-Poletti (eds), European Research Activities in Cloud Computing, CSP, 2012

•M. Rak et al,Security Issues in Cloud Federation.Achieving Federated&Self-Manageable Cloud Infrastructures,176-194, •F Moscato et al An Ontology for the Cloud in mOSAIC Cloud Computing: Methodology Systems&Applications 467-486F. Moscato et al, An Ontology for the Cloud in mOSAIC. Cloud Computing: Methodology,Systems&Applications,467 486 •D. Petcu et al, Towards Open-Source Cloudware, UCC 2011, 330-331.

•S. Panica et al,Serving Legacy Distributed Appls by a Self-configuring Cloud Processing Platform,IDAACS’11,139-145. •D. Petcu, Portability and Interoperability between Clouds: Challenges and Case Study, LNCS 6994, 62–74, 2011.

•D. Petcu et al, Building an Interoperability API for Sky Computing, InterCloud 2011, 405-412 •D. Petcu et al, Architecturing a Sky Computing Platform,, LNCS 6569, 2011, 1-13

•M. Frincu et al, Self-Healing Distributed Scheduling Platform, CCGrid'11, 225 - 234

•S. Venticinque etal,Agent based Cloud provisioning&management.Design&Prototypal Implementation,CLOSER’11,184-191 •D. Petcu et al, Towards a cross platform Cloud API. Components for Cloud Federation, CLOSER 2011, 166-169

•B. Di Martino et al, Building a Mosaic of Clouds, LNCS 6586, 2011, 529-536

•S. Venticinque et al, A Cloud Agency for SLA Negotiation and Management, LNCS 6586, 2011, 587-594

D P t t l F G id C ti T d Sk C ti C St d f E th Ob ti CGW 2010 11 20

23

•D. Petcu et al, From Grid Computing Towards Sky Computing. Case Study for Earth Observation, CGW 2010, 11-20 •D. Petcu, Identifying Cloud Computing Usage Patterns, Cluster 2010.

References

Related documents

Therefore, the notion of social and cultural capital as broader sociological concepts will aid in how their interplay with and through the habitus, agents negotiate

Uzziah, Menahem began to reign over Israel, and he reigned ten years ; 2 Kings xv.. Perhaps

We found that urban circular policies are positively and significantly associated with the attraction of greenfield FDI–generated employment, where the carrots (economic

Students will quickly review concepts from Math 6 and then proceed to learn and reinforce topics from Math 7 including: analyzing patterns, sequences, functions, and

• To install this utility, select the Prerequisite Software link on the SmartPlant Review product CD and select License Checkout Utility. The utility is installed in

Malik (1999) had done a limited research on advertisement effectiveness and consumer behavior of women university students but the present study intended to expand the sample in

While the low level of trade interdependence in ASEAN+3 is not immediately obvious when looking at intraregional trade share and trade intensity, we can conclude that regional

Unlike the acquaintance network which updates the agent list only during the addition of new agent, SCTs update the agent list whenever a transaction happens