Page 1
ResearchResearch
University of StuttgartUniversitätsstr. 3870569 StuttgartGermany
Phone +49-711-685 88337 Fax +49-711-685 88472
Santiago Gómez Sáez and Frank LeymannInstitute of Architecture of Application Systems
{gomez-saez,leymann}@iaas.uni-stuttgart.de
Design Support for Performance-aware Cloud
Application (Re-)Distribution
ESOCC PhD 2014
Page 2
Rese
arc
h
© Santiago Gómez Sáez 2
Agenda
Motivation & Problem Statement Related Works Research Challenges Work in Progress
Approach Experiments
Research Plan
Page 3
33© Santiago Gómez Sáez
Rese
arc
h
Motivation – Efficient Application Distribution
WebShop: WAR
Apache_Tomcat:Servlet_Container
Ubuntu10.04:Virt_Linux_OS
IBM_Server:Physical_Server
Product_DB:SQL_DB
MySQL: SQL_RDBMS_Server
AWS_EC2_m1.xlarge:
AWS_EC2
Ubuntu13.10:Virt_Linux_OS
MySQL: SQL_RDBMS_Server
AWS_RDS_mediumDB: AWS_RDS
MySQL: SQL_DBaaS
AWS_EC2_m1.medium:
AWS_EC2
Ubuntu13.0:Virt_Linux_OS
AWS_Elastic_BeansTalk: Application_Container
Partial vs. Complete Migration Multiple deployment options Multi-dimensional & Evolving problem Application workload behavior fluctuations Resources Demands Evolution
Performance-aware specification
Workload specification
alt_hosted_on
hosted_on
interacts-with
Andrikopoulos et al.: Optimal Distribution of Applications in the Cloud. In: Proceedings of CAiSE’14
Application specificAlternative node
Page 4
44© Santiago Gómez Sáez
Rese
arc
h
Motivation - Perspectives & Approaches
Performance-aware Specification
Workload Model Derivation & Characterization
Application Workload Evolution Monitoring
Top-down Bottom-up
Gómez Sáez et al.: Towards Dynamic Application Distribution Support for Performance Optimization in the Cloud. Proceedings of CLOUD 2014.
Page 5
55© Santiago Gómez Sáez
Rese
arc
hRelated Works
Cloud application design topology specification -> TOSCA, Blueprints, AWS Cloud Formation workload specification -> GT-CWSL
Cloud application design decision support MADCAT Palladio Component Model CloudMig MOCCA
Elasticity VM auto scaling techniques & algorithms proactive vs. reactive
Performance expectation vs. resource allocation specification
Page 6
66© Santiago Gómez Sáez
Rese
arc
hResearch Challenges
How to partially or completely specify during the design phase the Cloud Application topology and its performance-aware aspects?
Alternative topologies space derivation & pruning How to analyze the application workload behavior &
evolution towards deriving & assessing
an efficient application (re-)distribution & profitable resources configuration?
Page 7
77© Santiago Gómez Sáez
Rese
arc
h
Participants Application Developer
Application design and realization Application topology specification, e.g. using TOSCA
Application Cloud Distribution Design Support System Application (Re-)distribution towards proactively react to fluctuating workloads
Topology+ : enriched application topology model with performance awareness (e.g. expected throughput/operation or
component, resource consumption) application workload characteristics (e.g. probability matrix of operations)
Topology*: viable topology describing application distribution alternatives and specifying Cloud services dynamic resource adaptation configurations
WiP – Performance Aware Application (Re-)Distribution Process
Page 8
88© Santiago Gómez Sáez
Rese
arc
hWiP – Performance Aware Application (Re-)Distribution Process
Re-distribution
Model Applicatio
n Topology
Enrich Topology
Model
Derive WL Model & Topology Alternativ
es
Deployment &
Production
Evaluate (Re-)
Distribution
Performance Evolution
Performance
Registration
Monitor &
Analysis
Functional annotation
Performance-aware annotation
Legend
• Requirements• Capabilities• Constraints
• Expected Performance• Workload Behavior• Topology+
• Workload Model Derivation• Cloud Offerings Matching• Similarity & utility-based Analysis• Topologies*
• Topology* instance• Synthetic Workload Generation• Distribution Performance Evaluation
• Demanded vs. Provided Performance• Workload behavior Patterns• Statistical Classification
Collaborative Loop• Proactiveness• Workload Fluctuation• Application Performance Evolution• Optimize performance vs. cost
tradeoffGómez Sáez et al.: Towards Dynamic Application Distribution Support for Performance Optimization in the Cloud. Proceedings of CLOUD 2014.
Page 9
99© Santiago Gómez Sáez
Rese
arc
h
WebShop: WAR
…
Ubuntu10.04:Virt_Linux_OS
IBM_Server:Physical_Server
Product_DB:SQL_DB
MySQL: SQL_RDBMS_Server
AWS_EC2_m1.xlarge:
AWS_EC2
Ubuntu13.10:Virt_Linux_OS
AWS_RDS_xlargeDB: AWS_RDS
MySQL: SQL_DBaaS
alt_hosted_onhosted_on
WL Specification
Ubuntu10.04:Virt_Linux_OS
FlexiScale4vCPU:FlexiScale_VM
on-premise
Interacts-with
IaaS DBaaS
WiP - Experiments
Gómez Sáez et al.: Towards Dynamic Application Distribution Support for Performance Optimization in the Cloud. Proceedings of CLOUD 2014.
TPC-H Benchmark as the basis Workload model derivation Different workload characteristics
Page 10
1010© Santiago Gómez Sáez
Rese
arc
hWiP – Experimental Results
LegendCL: Compute LowCM: Compute MediumCH: Compute High
Page 11
1111© Santiago Gómez Sáez
Rese
arc
h
Research Plan
Flesh out the individual process tasks application topology specification application topology enrichment application workload analysis & generation relationship between developer preferences & application performance …
Performance experiments on application upper layers WordPress MediaWiki
Performance evaluation & analysis of the overall process distribution vs. redistribution re-configuration Santiago Gómez Sáez
E-mail: [email protected] of Architecture of Applications Systems (IAAS)University of Stuttgart (Germany)
Page 12
12
Research
Thanks for your attention!!