• No results found

OPTIMIS: Improving Cloud Management With Dynamic SLAs

N/A
N/A
Protected

Academic year: 2021

Share "OPTIMIS: Improving Cloud Management With Dynamic SLAs"

Copied!
12
0
0

Loading.... (view fulltext now)

Full text

(1)

1  

OPTIMIS:  Improving  Cloud  

Management  With  Dynamic  SLAs  

Salt  Lake  City,  July  18  

 

Wolfgang  Ziegler,  Fraunhofer  Institute  SCAI  

(2)

2  

OPTIMIS  Project  

IP  5th  call  

June  2010  -­‐  May  2013  

13  Partner  

Atos  Origin  SAE  

Umeå  University  

The  451  Group  

Universität  Stuttgart  

National  Technical  University  of  Athens  

Barcelona  Supercomputing  Center  

SAP  

Fraunhofer  Institute  for  Algorithms  and  Scientific  Computing  

University  of  Leeds  

Leibniz  University  of  Hanover  

Flexiant  

British  Telecom  

Arsys  

(3)

3  

OPTIMIS  Scenarios  

(4)

4  

SLA  use  in  OPTIMIS  

Provide  SLA-­‐aware  infrastructure  services  with  extended  

QoS  capabilities  

SLAs  between  infrastructure  providers  and  service  providers  

Support  for  self-­‐*  and  elasticity  

Definition  and  standardisation  of  new  term  languages  

Trust  

Risk  

Eco-­‐efficiency  

Cost  

Data  security  &  legal  constraints  

(5)

5  

OPTIMIS  SLA  Management  

Cloud  QoS  

 

OPTIMIS  SLA  Management

 

 

Standard  compliant  SLA  Management  component

 

 

Full  support  of  the  OGF  WS-­‐Agreement    specification

 

 

SLA  Negotiation  supported  through  WS-­‐Agreement  Negotiation

 

 

Based  of  the  WSAG4J  SLA  Management  Framework  developed  at  Fraunhofer  SCAI

 

 

Functionality  of  the  SLA  Management  Layer

 

 

Mediates  between  SDO  and  IP  components

 

 

Negotiates  IP  services  and  triggers  deployment

 

 

Basic  integration  with  IP  components  for  service  deployment  implemented

 

 

Integration  of  SLA  monitoring  planned  for  year  2

 

 

Innovations

 

 

Support  of  TREC  parameters  as  part  of  the  SLA

 

  Enables  self-­‐management  and  self-­‐optimization  operations  based  on  TREC  parameters

 

 

Dynamic  SLA  creation,  monitoring  and  accounting

 

  Rule-­‐based  definition  of  SLA  guarantees

 

 

Dynamic  detection  of  SLA  compliance  and  violations

 

 

Trigger  self-­‐management  and  self-­‐optimization  actions  in  order  to  prevent  SLA  violations

 

(6)

SLA  Management  

SLA  Creation  

SLA  Validation  

 

Template  based  creation  of  SLAs  

 

Templates  contain  restrictions  on  valid  

agreement  offers  

  Structural  restrictions,  i.e.  which  element  

must  be  contained  in  the  offer  and  how   often  

  Value  restrictions,  i.e.  what  are  valid  values  

for  specific  elements  

 

Enforcement  of  offer  integrity  

  Constraint  validation  before  service  

instantiation  

  Protection  of  IP  components  

Service  Instantiation  

 

Admission  of  services  done  by  IP  admission  

control  component  

 

Deployment  agreed  services  initiated  by  IP  

cloud  optimizer  

Agreement  Factory   Facade   Create  Agreement   Agreement  Factory   Resource   Proper5es   Templat es Service   Provisioning   System  

Infrastructure Provider SLA Management (Cloud QoS)

GetTemplates Get  Templates  

Templates Service Deployment Optimizer Agreement   Offer   Valida5on   Service   Instan5a5on   Agreement   Offer   Validator   Agreement   Crea5on   Create Agreement Agreement EPR Lookup Lookup Validate Instantiate   Configura5on   Populate Configure

(7)

SLA  Management  

Monitoring  

SLA  Monitoring  

 

Periodical  monitoring  of  deployed  

services  and  guarantees  

 

Guarantee  states  are  computed  based  

on  definition  in  SLA

 

 

Guarantees  can  specify  either  an  

importance  or  monetary  values  

 

Penalties  and  rewards  with  monetary  

values  result  in  notifications  to  an  

accounting  system

 

 

Penalties  and  rewards  with  an  

importance  result  in  notifications  to  

the  self-­‐management  and  self-­‐

optimization  components  

Agreement   Monitor   Agreement Guarantee   Evaluator   Agreement Properties D yn am ic R es o u rc e Pr o p er ti es Sta ti c R es o u rc e Pr o p er ti es Terms Service Property 2 Service Property 1 Guarantee Term

Service Term States Service Term State 1

State

Guarantee Term States Guarantee Term

State Agreement State

...

Service Description Term 1

Monitoring  System   Accoun5ng   compute states 1. retrieve service states Self-­‐Management  /   Self  Op5miza5on   2. update service terms states 4. update guarantee states 3. evaluate guarantees 5. notify

(8)

SLA  Management  

Monitoring  of  SLAs  

Static  Properties  

 

Service  Description  

  Specify  the  user  requirements  on  the  service   provisioning  process  

  Can  be  changed  by  the  user  with  respect  to  the   agreement  creation  constraints  

 

Service  Properties  

  Define  variables  in  order  to  resolve  static  and   dynamic  property  values  

  Are  used  to  define  the  service  level  objectives  in  SLA   guarantees  

 

Guarantees  

  Specify  service  level  objectives  

  Define  self  management  events  and  accounting   models  in  form  of  penalties  and  rewards  

Dynamic  Properties  

 

Service  Term  States    

  Dynamically  generated  during  SLA  monitoring     Contain  detailed  information  on  the  current  service  

delivery  

 

Guarantee  Term  States  

  Are  automatically  evaluated  by  the  SLA  management     Accounting  and  Self-­‐management  capabilities  are  

triggered  based  on  evaluation  results   Agreement D yn am ic Pr o p er ti es Sta ti c Pr o p er ti es Terms Service Property 2 Service Property 1 Guarantee

Service Term States

Service Term State 1

State

Guarantee Term States

Guarantee Term State Agreement State ... Service Description agreed service property actual service property

(9)

9  

(10)

10  

OPTIMIS  Service  Manifest  

 

Describes  the  SP  view  on  

the  Services

 

 

OVF  XML  template

 

 

Base  for  the  Service  

Description  Terms  of  the  

Service  Level  

Agreement  between  SP  

and  IP

 

 

Each  section  in  the  OVF  

template  is  mapped  to  

one  of  the  Service  

Description  Terms  in  the  

Agreement  template  

<opt:ElasticitySection> <opt:Rule> <opt:Scope> <opt:VirtualSystemGroupId>instance-jboss-group</opt:VirtualSystemGroupId> </opt:Scope>

<!-- Current amount of JVM Threads in the JBoss JVM --> <opt:KPIName>ThreadCount</opt:KPIName>

<!-- These values are not used for this measurement --> <opt:Window>P5M</opt:Window>

<opt:Frequency>1</opt:Frequency>

<!-- 100 users per thread, with a tolerance of 5% (so trigger at more than 105 users, and down again at below 95 users ) -->

<opt:Quota>100</opt:Quota> <opt:Tolerance>5</opt:Tolerance> </opt:Rule> </opt:ElasticitySection> <opt:DataProtectionSection> <opt:DataProtectionLevel>DPA</opt:DataProtectionLevel> <opt:DataEncryptionLevel> <opt:EncryptionAlgoritm>AES</opt:EncryptionAlgoritm> </opt:DataEncryptionLevel> </opt:DataProtectionSection>

<ws:ServiceDescriptionTerm ws:Name="TREC" ws:ServiceName="MultipleImages"> <opt:TRECSection> <opt:TrustSection> <opt:TrustLevel>5</opt:TrustLevel> </opt:TrustSection> <opt:RiskSection> <opt:AvailabilityArray>

<!-- defines a minimum availability of the VM of 98% at a day --> <opt:Availability opt:AssessmentInterval="P1D">98</opt:Availability> <!-- defines a minimum availability of the VM of 99% at a month --> <opt:Availability opt:AssessmentInterval="P1M">99</opt:Availability> </opt:AvailabilityArray> </opt:RiskSection> <opt:EcoEfficiencySection> <opt:LEEDCertification>NotRequired</opt:LEEDCertification> <opt:BREEAMCertification>NotRequired</opt:BREEAMCertification> <opt:EuCoCCompliant>false</opt:EuCoCCompliant> <opt:EnergyStarRating>No</opt:EnergyStarRating> </opt:EcoEfficiencySection> <opt:CostSection> <opt:Price opt:currency="EUR">100.00</opt:Price> </opt:CostSection> </opt:TRECSection> </ws:ServiceDescriptionTerm>‚

(11)

11  

(12)

12  

Service  Details  Examples  (TREC,  Elasticity,  

Data  Protection)  

Trust  

TrustLevel  5  

Risk  

minimum  availability  of  the  VM  of  99%  at  a  month  

Eco  Efficiency  

No  certification  required  

Cost  

100.00  Euro  (maximum  cost)    

Elasticity  (jboss  instances)  

100  users  per  thread,  with  a  tolerance  of  5%  (so  trigger  at  more  

than  105  users,  and  down  again  at  below  95  users  

Data  Protection  

Data  Protection  Level  DPA  

References

Related documents

We also derive a Feynman–Kac formula for the stochastic heat equation in the Skorokhod sense and we obtain the Wiener chaos expansion of the solution.. Fractional noise, stochastic

02 Integrate DM into school curriculum to raise awareness among students on disaster preparedness, mitigation and response Develop students curriculum to integrate DM

PCS-931 also includes distance protection (3 forward zones and 1 reverse zone distance protection with selectable mho or quadrilateral characteristic), 4 stages

Lukenya Notes: Pro-poor Sanitation Marketing and Sustainability beyond ODF September 2011 Programme (to be launched in 2011). This involves i) market research, ii)

The CDC country office coordinates with relevant Ministries’ leadership to explain the multi-tiered approach to health systems strengthening, determine priority health systems

Sourcing Management and Execution Models Client NAI Service Provider 1 Service Provider 2 Service Provider n • Service Fees • Contracts – SLAs • Process Interactions • Service

SLM processes provide a framework by which services are defined, levels of service required to support business processes agreed upon, Service Level Agreements (SLAs)

Previous literature provides possible explanations for these mixed findings, including (i) the possible inexistence of a sequenced set of activities designed to achieve