Upload
daniel-moldovan
View
159
Download
0
Tags:
Embed Size (px)
DESCRIPTION
Presentation given in PhD Session of advanced School on Service Oriented Computing 30 June – 5 July, 2014, Hersonissos Crete Greece (http://www.summersoc.eu/program/) Contains in last slide embedded video of the given talk. Overview of techniques for analyzing elasticity of both cloud offered services and running cloud services, towards supporting the design, and control of elastic cloud services.
Citation preview
1
Novel Models and Techniques for Monitoring and Analysis of
Software-defined Elastic SystemsDaniel Moldovan
Distributed Systems Group, Vienna University of Technology
[email protected]://www.infosys.tuwien.ac.at/staff/dmoldovan/
SummerSoc 2014
SummerSoc 2014 2
Agenda
• Analyzing Elasticity of cloud offered services• Quantifying elasticity of cloud offered services
• Monitoring elastic cloud services• Analyzing Elasticity of cloud services
• Elasticity Space, Boundaries and Pathway• Elasticity Dependencies
Analyzing Elasticity of cloud offered servicesMotivation
SummerSoc 2014 3
What cloud offered service provides appropriate elasticity for me?
Analyzing Elasticity of cloud offered servicesMotivation
Load balancer
Cost: /H & / GB
SummerSoc 2014 4
For non-elastic components easier to select service
Analyzing Elasticity of cloud offered servicesMotivation
Load balancer ...
Cost: /H & / GB
?
SummerSoc 2014 5
What cloud offered service provides appropriate elasticity?
Analyzing Elasticity of cloud offered servicesMotivation
Load balancer
t1.micro EBS+
...
Cost: /H & / GB
?
...
Cost: - OnDemand- Spot- Reserved (1/3 years)
Cost: /size/month
SummerSoc 2014 6
Should I select one service with higher Cost elasticity (more cost options), but with a mandatory dependency?
Analyzing Elasticity of cloud offered servicesMotivation
Load balancer
t1.micro EBS+ i2.xlarge EBS+
...
Cost: /H & / GB
?
...
Cost: - OnDemand- Spot- Reserved (1/3 years)
Cost: /size/month
Cost: - OnDemand- Reserved (1/3 years)
Cost: /size/month
SummerSoc 2014 7
Should I go with a service with less Cost elasticity but an optional dependency on another service?
Analyzing Elasticity of cloud offered services Approach
SummerSoc 2014 8
Elasticity Capability Elasticity Dependencies
Service Unit
Analyzing Elasticity of cloud offered servicesModelling Elasticity Capabilities
Costm1.tiny
OnDemand
Reserved 1Y light
Spot
...SummerSoc 2014 9
Analyzing Elasticity of cloud offered servicesCustomizable Elasticity Quantification Coefficients
• Elasticity Phase Quantification Coefficients
• Elasticity Dependency Type Quantification Coefficients
• Elasticity Dependency Volatility Quantification Coefficients
• Elasticity Capability Quantification Function
•
• Elasticity Quantification Function
SummerSoc 2014 10
Analyzing Elasticity of cloud offered servicesExample: Elasticity of Amazon EC2 IaaS Services
SummerSoc 2014 11
• Elasticity Phase Quantification Coefficient
• Elasticity Dependency Volatility Quantification Coefficient
• Elasticity Dependency Type Quantification Coefficient
SummerSoc 2014 12
Analyzing Elasticity of cloud offered services
SummerSoc 2014 13
• Analyzing Elasticity of cloud offered services• Quantifying elasticity of cloud offered services
• Monitoring elastic cloud services• Analyzing Elasticity of cloud services
• Elasticity Space, Boundaries and Pathway• Elasticity Dependencies
Monitoring elastic cloud servicesService Level Monitoring
Response time
Number of clients
Other specific metrics
System Level MonitoringGanglia, Nagios, etc.
CPU usage
Memory usage
Network transfer
14
User-Defined Requirements violation: - Cost per client too highReasons: - Too much logging? Monitoring chatter? - Too expensive VMs? Which one can be downsized? - Not enough clients? Why?
Controlling the service’s elasticity
SummerSoc 2014
Daniel Moldovan, Georgiana Copil, Hong-Linh Truong, Schahram Dustdar, MELA: Monitoring and Analyzing Elasticity of Cloud Services, 5'th International Conference on Cloud Computing, CloudCom. Bristol, UK, 2-5 December, 2013, http://dx.doi.org/10.1109/CloudCom.2013.18
Monitoring elastic cloud servicesStructuring monitoring information
15
Monitoring Data
Service Structure
SummerSoc 2014
Monitoring elastic cloud servicesStructuring monitoring information
16
Monitoring Data
Service Structure
Impose service structure over collected monitoring data
SummerSoc 2014
Monitoring elastic cloud servicesStructuring and enriching monitoring information
17SummerSoc 2014
Compute number of used virtual machines per unit
Monitoring elastic cloud servicesStructuring and enriching monitoring information
18SummerSoc 2014
Compute number of used virtual machines per unit
Monitoring elastic cloud servicesStructuring and enriching monitoring information
19SummerSoc 2014
Enrich metric with COST information COST/VM * numberOfVMs
Monitoring elastic cloud servicesStructuring and enriching monitoring information
20SummerSoc 2014
Propagate activeConnections from LoadBalancer service unit
Monitoring elastic cloud servicesStructuring and enriching monitoring information
21SummerSoc 2014
Compute cost/client/h
SummerSoc 2014 22
• Analyzing Elasticity of cloud offered services• Quantifying elasticity of cloud offered services
• Monitoring elastic cloud services• Analyzing Elasticity of cloud services
• Elasticity Space, Boundaries and Pathway• Elasticity Dependencies
SummerSoc 2014 23
Data Controller
Event Processing
clientsLoad Balancer Data Node
Analyzing elasticity of cloud servicesMotivation
Elastic Service
User-Defined Elasticity Requirements
Cost <= 2.5 $/Month/Client Response Time <= 1 s
Daniel Moldovan, Georgiana Copil, Hong-Linh Truong, Schahram Dustdar, MELA: Monitoring and Analyzing Elasticity of Cloud Services, 5'th International Conference on Cloud Computing, CloudCom. Bristol, UK, 2-5 December, 2013, http://dx.doi.org/10.1109/CloudCom.2013.18
SummerSoc 2014 24
Data Controller
Event Processing
clientsLoad Balancer Data Node
Analyzing elasticity of cloud servicesMotivation
Elastic Service
• When to scale Event Processing? • Requirements for Event Processing?
User-Defined Elasticity Requirements
Cost <= 2.5 $/Month/Client Response Time <= 1 s
Daniel Moldovan, Georgiana Copil, Hong-Linh Truong, Schahram Dustdar, MELA: Monitoring and Analyzing Elasticity of Cloud Services, 5'th International Conference on Cloud Computing, CloudCom. Bristol, UK, 2-5 December, 2013, http://dx.doi.org/10.1109/CloudCom.2013.18
SummerSoc 2014 25
Data Controller
Event Processing
clientsLoad Balancer Data Node
Analyzing elasticity of cloud servicesMotivation
Elastic Service
• When to scale Event Processing? • Requirements for Event Processing?
• When to scale Data Node? • Requirements for Data Node?
User-Defined Elasticity Requirements
Cost <= 2.5 $/Month/Client Response Time <= 1 s
Daniel Moldovan, Georgiana Copil, Hong-Linh Truong, Schahram Dustdar, MELA: Monitoring and Analyzing Elasticity of Cloud Services, 5'th International Conference on Cloud Computing, CloudCom. Bristol, UK, 2-5 December, 2013, http://dx.doi.org/10.1109/CloudCom.2013.18
26
User-Defined Elasticity Requirements
Cost <= 2.5 $/Month/Client Response Time <= 1 s
SummerSoc 2014
Analyzing elasticity of cloud servicesElasticity Space, Boundaries and Pathway
Daniel Moldovan, Georgiana Copil, Hong-Linh Truong, Schahram Dustdar, MELA: Monitoring and Analyzing Elasticity of Cloud Services, 5'th International Conference on Cloud Computing, CloudCom. Bristol, UK, 2-5 December, 2013, http://dx.doi.org/10.1109/CloudCom.2013.18
27
User-Defined Elasticity Requirements
Cost <= 2.5 $/Month/Client Response Time <= 1 s
SummerSoc 2014
Analyzing elasticity of cloud servicesElasticity Space, Boundaries and Pathway
Daniel Moldovan, Georgiana Copil, Hong-Linh Truong, Schahram Dustdar, MELA: Monitoring and Analyzing Elasticity of Cloud Services, 5'th International Conference on Cloud Computing, CloudCom. Bristol, UK, 2-5 December, 2013, http://dx.doi.org/10.1109/CloudCom.2013.18
Elasticity Space Snapshot
Elasticity Space “Clients/h” Dimension
Elasticity Space “Response Time” Dimension
Analyzing elasticity of cloud servicesElasticity Space and Boundary
User-Defined Elasticity Requirements
Cost <= 2.5 $/Month/Client Response Time <= 1 s
SummerSoc 2014 28
Elasticity Space Snapshot
Elasticity Space “Clients/h” Dimension
Elasticity Space “Response Time” Dimension
Analyzing elasticity of cloud servicesElasticity Space and Boundary
User-Defined Elasticity Requirements
Cost <= 2.5 $/Month/Client Response Time <= 1 s
SummerSoc 2014 29
Event Processing service unit Elasticity Pathway
Analyzing elasticity of cloud servicesElasticity Pathway
30SummerSoc 2014
SummerSoc 2014 31
Agenda
• Analyzing Elasticity of cloud offered services• Quantifying elasticity of cloud offered services
• Monitoring elastic cloud services• Analyzing Elasticity of cloud services
• Elasticity Space, Boundaries and Pathway• Elasticity Dependencies
Analyzing elasticity of cloud servicesElasticity Dependencies
32SummerSoc 2014
Enforce Elasticity Capability (scale in/our, reconfigure)
Effect ?
Change in Load
Conclusions• Concepts
• Elasticity Space, Boundary, Pathway
• Elasticity Dependencies
• Elasticity Capabilities
• Mechanisms• Quantifying elasticity of cloud offered services
• Constructing cross-layer monitoring snapshots
• Analyzing elasticity
• Framework/Tools• QUELLE: a Framework for Accelerating the Development of Elastic Systems (
http://tuwiendsg.github.io/QUELLE)
• MELA: Monitoring and analyzing elasticity of cloud services (http://tuwiendsg.github.io/MELA/)
Work partially supported by the European Commission in terms of the CELAR FP7 project (http://www.celarcloud.eu/)
Distributed Systems Group (http://dsg.tuwien.ac.at/)
Vienna University of Technology (http://www.tuwien.ac.at/)
SummerSoc 2014 33