SlideShare una empresa de Scribd logo
1 de 17
Descargar para leer sin conexión
Ming Mao, Marty Humphrey
CS Department, UVa
Scaling and Scheduling to Maximize
Application Performance within Budget
Constraints in Cloud Workflows
IPDPS 2013
(May 21st 2013)
1
2
 Dynamic scalability and cost saving are two of the most important factors when
considering cloud adoption
Two major benefits
- dynamic scalability and cost
A survey from 39 major technology companies [1]
 Cloud benefits
 On-demand self-services
 Broad network access
 Resource pooling
 Rapid elasticity
 Measured services
 Cheaper maintenance
 ……
Why do you move into the cloud?
3
 Dynamic scalability – the ability to acquire/release resources in response to
demand dynamically
 Dynamic scalability challenge → It relies on the users to tell the size of resource
pool
 Over-provisioning → cost more than necessary, offset cloud advantages
 Under-provisioning → hurt application performance, cannot meet service level agreements and
lose application customers
Cloud dynamic scalability
over-provisioning under-provisioning
4
 Problem - What resources should be acquired/released in the cloud,
and how should the computing activities be mapped to the cloud
resources, so that the application performance can be maximized
within the budget constrains?
 In this paper, we discuss limited budget case
 The unlimited budget case was discussed in our SC 11 paper
 Solution - This paper argues that an automatic resource
provisioning and allocation mechanism, i.e., an auto-scaling
solution – is the key to successful cloud adoption. Essentially, an
auto-scaling solution needs to answer the following two questions:
 Capacity determination (or resource provisioning)
 what types of resources, how much and for how long
 Job scheduling (or resource allocation)
 map computing activities onto the cloud resources
Problem statement
5
 An application consists of service components. A workflow goes through different
service components and therefore consists of multiple connected tasks
 Workload is a stream of workflow jobs not known in advance
 Task precedence constraints need to be preserved
 Jobs have individual priorities
Service oriented architecture (SOA) & workflow jobs
6
Minimize job turnaround time within budget constraints
Problem formulation
 Problem terminology
 Cloud application
 app = {Si}
 Job class
 J = {DAG(Si), priorityJ| Si ∈ app}
 Cloud VM
 VMv = {[𝐽 𝑆 𝑖]v , cv , lagv}
 Workload
 Wt = 𝑗𝑜𝑏𝐽
𝑆 𝑖
𝑗𝑜𝑏𝐽𝑆 𝑖
 Scaling plan
 Scalingt = {VMv → Nv}
 Scheduling plan
 Schedulet = { 𝑗𝐽
𝑆 𝑖
→VMv}
 Goal
 Min( 𝑗𝑜𝑏𝑡𝑢𝑟𝑛𝑎𝑟𝑜𝑢𝑛𝑑 × 𝑝𝑟𝑖𝑜𝑟𝑖𝑡𝑦/𝑗𝑜𝑏 𝑝𝑟𝑖𝑜𝑟𝑖𝑡𝑦𝑗𝑜𝑏 )
&&
 Cost(app) <= B (budget, dollars/hour)
 Target - The service provider has a limited budget and
aims to maximize the application performance.
 Solution idea – a monitor-control loop that
makes scaling and scheduling decisions based
on updated workload and VM information
7
 Scheduling-first
 Idea – allocate application budget to individual jobs based on priorities
and schedule tasks within job budget
 Step 1 – Distribute budget: 𝐵𝑗 = 𝐵 × 𝑝𝑗/ 𝑝𝑗𝑗
 Step 2 – Schedule tasks
 for each job, schedule as many tasks as possible on their fast machines
 Step 3 – Consolidate budget
 return job budget to the application
 the application uses the remaining budget collected from individual jobs to schedule
high priority tasks
 Step 4 – Acquire instance
 acquire instances and execute tasks based on the determined schedule plans
Minimize job turnaround time within budget constraints
Solution: scheduling-first
8
 Scheduling-first
 Step 1 – Distribute budget: 𝐵𝑗 = 𝐵 × 𝑝𝑗/ 𝑝𝑗𝑗
Minimize job turnaround time within budget constraints
Solution: scheduling-first
 Step 2 – Schedule tasks
e.g. Budget(B) = $1/h;
Large(L) = $0.5/h;
Medium(M) = $0.3/h;
Small(S) = $0.1/h;
 Step 1: job1 and job2 have
the same priority,
job1 → $0.5/h, job2 → $0.5/h
 Step 2:
job1(T1) → $0.5(L);
job2(T5) → $0.5(L);
 Step 3:
job1(T2+T3) → $0.5(S+M);
job2(T6) → $0.5(L);
job1 returns $0.1 to system;
job2(T7) → $0.1(S);
 Step 4
acquire instances when
necessary
 Step 3 – Consolidate budget
 Step 4 – Acquire instance
9
Minimize job turnaround time within budget constraints
Solution: scaling-first
 Scaling-first
 Idea – determine the computing capacity by looking at the overall
workload and schedule tasks based on priority
 Step 1 – determine the VMs
 assume tasks run on their fastest machines and calculate the cost Cfast for the next
hour
 acquire VMs proportionally based on Budget/Cfast
 Step 2 – consolidate budget
 use the remaining the budget to purchase new machines.
 Step 3 – schedule tasks
 schedule tasks based on task priority
10
Minimize job turnaround time within budget constraints
Solution: scaling-first
 Scaling-first
 Step 1 – determine the VMs
 Step 2 – consolidate budget
 Step 3 – schedule tasks
 Step 1: assume tasks run on fastest
machines and calculate Cfast and
acquire VMs proportionally based on
B/Cfast,
 Step 2: the remaining $0.5 can be used to
purchase 1 L machine
 Step 3: tasks are scheduled
based on their priorities
11
 Instance consolidation
 Schedule tasks on different VM types to save partial instance hour cost
 Budget allocation schemes
 Evenly distributed – e.g. daily x/365, hourly x/8760
 Based on workload – e.g. high on busy times, low on non-busy times
 Workload prediction – $/hour → $/job
Minimize job turnaround time within budget constraints
Other considerations
 Workload patterns
 Application models
12
 Time
 72 hours
 Task execution
 Randomly generated
 VM lag
 5 min
Minimize job turnaround time within budget constraints
Evaluation – experiment setup
 Baseline
 Standard
VM Type Price
Micro $0.02/hour
Standard $0.080/hour
High-CPU $0.66/hour
High-Memory $0.45/hour
Extra-Large $1.3/hour
13
Minimize job turnaround time within budget constraints
Evaluation – job turnaround time
 above – weighted average job turnaround time for the hybrid application and cycle
workload pattern
 Scheduling-first and scaling-first can save 9.8%- 45.2% cost compared to the standard
machine choice.
 Scaling-first works better under small budget ranges while scheduling-first works better
under large budget ranges.
14
Minimize job turnaround time within budget constraints
Evaluation – sensitivity to inaccurate parameters
 left – scheduling-first’s sensitivity to inaccurate parameters (Hybrid application + Cycle
workload pattern)
 right – scaling-first’s sensitivity to inaccurate parameters (Hybrid application + Cycle workload
pattern)
 When the estimation error is within ±20%, the job turnaround time shows -10.2% – 16.7%
difference.
 When the task estimation error reaches ±60%, the performance of both algorithms shows
significant degradation (more than ±25% difference)
15
Minimize job turnaround time within budget constraints
Evaluation – instance consolidation
 left – job turnaround time / resource utilization of scheduling-first’s instance consolidation
(Hybrid application + Cycle workload pattern)
 right – job turnaround time / resource utilization of scaling-first’s instance consolidation
(Hybrid application + Cycle workload pattern)
 When budget is low or high, the improvement is small. When the budget is in between, the
improvement is more significant (e.g. utilization rate improves 2.2% to 19.9% when the budget
is between $15/hour and $25/hour).
 Scaling-first benefits more from instance consolidation process than scheduling-first
16
 Conclusions
 choose appropriate VM types based on the workload.
 Scheduling-first and scaling-first are trade-offs between the task execution time and
waiting time.
 As long as the VM performance can be correctly ranked, the proposed mechanisms have
good tolerance to inaccurate parameters.
 Instance consolidation is an efficient strategy to save partial instance hours and improve
resource utilization.
 Future work
 Other billing models – reserved instances, spot instances, $/min
 Maximize application performance within budget constraints for data-intensive
applications
 Hybrid and federate cloud environments
 Develop evaluation benchmarks and simulation platforms
Conclusion and future work
17
Thanks!

Más contenido relacionado

La actualidad más candente

Scheduling production
Scheduling   productionScheduling   production
Scheduling productionAditya Powale
 
IRJET- Advance Approach for Load Balancing in Cloud Computing using (HMSO) Hy...
IRJET- Advance Approach for Load Balancing in Cloud Computing using (HMSO) Hy...IRJET- Advance Approach for Load Balancing in Cloud Computing using (HMSO) Hy...
IRJET- Advance Approach for Load Balancing in Cloud Computing using (HMSO) Hy...IRJET Journal
 
Resource Scheduling and Evaluation of Heuristics with Resource Reservation in...
Resource Scheduling and Evaluation of Heuristics with Resource Reservation in...Resource Scheduling and Evaluation of Heuristics with Resource Reservation in...
Resource Scheduling and Evaluation of Heuristics with Resource Reservation in...Eswar Publications
 
REAL-TIME ADAPTIVE ENERGY-SCHEDULING ALGORITHM FOR VIRTUALIZED CLOUD COMPUTING
REAL-TIME ADAPTIVE ENERGY-SCHEDULING ALGORITHM FOR VIRTUALIZED CLOUD COMPUTINGREAL-TIME ADAPTIVE ENERGY-SCHEDULING ALGORITHM FOR VIRTUALIZED CLOUD COMPUTING
REAL-TIME ADAPTIVE ENERGY-SCHEDULING ALGORITHM FOR VIRTUALIZED CLOUD COMPUTINGijdpsjournal
 
Psdot 1 optimization of resource provisioning cost in cloud computing
Psdot 1 optimization of resource provisioning cost in cloud computingPsdot 1 optimization of resource provisioning cost in cloud computing
Psdot 1 optimization of resource provisioning cost in cloud computingZTech Proje
 
Earning Value from Earned Value Management
Earning Value from Earned Value ManagementEarning Value from Earned Value Management
Earning Value from Earned Value ManagementGlen Alleman
 
Design of Capability Delivery Adjustments @ASDENCA
Design of Capability Delivery Adjustments @ASDENCADesign of Capability Delivery Adjustments @ASDENCA
Design of Capability Delivery Adjustments @ASDENCAJānis Grabis
 

La actualidad más candente (9)

operations scheduling
operations schedulingoperations scheduling
operations scheduling
 
Job shop scheduling problem using genetic algorithm
Job shop scheduling problem using genetic algorithmJob shop scheduling problem using genetic algorithm
Job shop scheduling problem using genetic algorithm
 
Scheduling production
Scheduling   productionScheduling   production
Scheduling production
 
IRJET- Advance Approach for Load Balancing in Cloud Computing using (HMSO) Hy...
IRJET- Advance Approach for Load Balancing in Cloud Computing using (HMSO) Hy...IRJET- Advance Approach for Load Balancing in Cloud Computing using (HMSO) Hy...
IRJET- Advance Approach for Load Balancing in Cloud Computing using (HMSO) Hy...
 
Resource Scheduling and Evaluation of Heuristics with Resource Reservation in...
Resource Scheduling and Evaluation of Heuristics with Resource Reservation in...Resource Scheduling and Evaluation of Heuristics with Resource Reservation in...
Resource Scheduling and Evaluation of Heuristics with Resource Reservation in...
 
REAL-TIME ADAPTIVE ENERGY-SCHEDULING ALGORITHM FOR VIRTUALIZED CLOUD COMPUTING
REAL-TIME ADAPTIVE ENERGY-SCHEDULING ALGORITHM FOR VIRTUALIZED CLOUD COMPUTINGREAL-TIME ADAPTIVE ENERGY-SCHEDULING ALGORITHM FOR VIRTUALIZED CLOUD COMPUTING
REAL-TIME ADAPTIVE ENERGY-SCHEDULING ALGORITHM FOR VIRTUALIZED CLOUD COMPUTING
 
Psdot 1 optimization of resource provisioning cost in cloud computing
Psdot 1 optimization of resource provisioning cost in cloud computingPsdot 1 optimization of resource provisioning cost in cloud computing
Psdot 1 optimization of resource provisioning cost in cloud computing
 
Earning Value from Earned Value Management
Earning Value from Earned Value ManagementEarning Value from Earned Value Management
Earning Value from Earned Value Management
 
Design of Capability Delivery Adjustments @ASDENCA
Design of Capability Delivery Adjustments @ASDENCADesign of Capability Delivery Adjustments @ASDENCA
Design of Capability Delivery Adjustments @ASDENCA
 

Destacado

Cloud auto-scaling with deadline and budget constraints
Cloud auto-scaling with deadline and budget constraintsCloud auto-scaling with deadline and budget constraints
Cloud auto-scaling with deadline and budget constraintsmingtemp
 
An Integrated Framework for Parameter-based Optimization of Scientific Workflows
An Integrated Framework for Parameter-based Optimization of Scientific WorkflowsAn Integrated Framework for Parameter-based Optimization of Scientific Workflows
An Integrated Framework for Parameter-based Optimization of Scientific Workflowsvijayskumar
 
Energy efficient resource management for high-performance clusters
Energy efficient resource management for high-performance clustersEnergy efficient resource management for high-performance clusters
Energy efficient resource management for high-performance clustersXiao Qin
 
A performance study on the vm startup time in the cloud
A performance study on the vm startup time in the cloudA performance study on the vm startup time in the cloud
A performance study on the vm startup time in the cloudmingtemp
 
Auto-Scaling to Minimize Cost and Meet Application Deadlines in Cloud Workflows
Auto-Scaling to Minimize Cost and Meet Application Deadlines in Cloud WorkflowsAuto-Scaling to Minimize Cost and Meet Application Deadlines in Cloud Workflows
Auto-Scaling to Minimize Cost and Meet Application Deadlines in Cloud Workflowsmingtemp
 
Ch5 - Project Management
Ch5 - Project ManagementCh5 - Project Management
Ch5 - Project ManagementJomel Penalba
 
Task Scheduling and Asynchronous Processing Evolved. Zend Server Job Queue
Task Scheduling and Asynchronous Processing Evolved. Zend Server Job QueueTask Scheduling and Asynchronous Processing Evolved. Zend Server Job Queue
Task Scheduling and Asynchronous Processing Evolved. Zend Server Job QueueSam Hennessy
 
Differentiating Algorithms of Cloud Task Scheduling Based on various Parameters
Differentiating Algorithms of Cloud Task Scheduling Based on various ParametersDifferentiating Algorithms of Cloud Task Scheduling Based on various Parameters
Differentiating Algorithms of Cloud Task Scheduling Based on various Parametersiosrjce
 
An Introduction To Applied Evolutionary Meta Heuristics
An Introduction To Applied Evolutionary Meta HeuristicsAn Introduction To Applied Evolutionary Meta Heuristics
An Introduction To Applied Evolutionary Meta Heuristicsbiofractal
 

Destacado (9)

Cloud auto-scaling with deadline and budget constraints
Cloud auto-scaling with deadline and budget constraintsCloud auto-scaling with deadline and budget constraints
Cloud auto-scaling with deadline and budget constraints
 
An Integrated Framework for Parameter-based Optimization of Scientific Workflows
An Integrated Framework for Parameter-based Optimization of Scientific WorkflowsAn Integrated Framework for Parameter-based Optimization of Scientific Workflows
An Integrated Framework for Parameter-based Optimization of Scientific Workflows
 
Energy efficient resource management for high-performance clusters
Energy efficient resource management for high-performance clustersEnergy efficient resource management for high-performance clusters
Energy efficient resource management for high-performance clusters
 
A performance study on the vm startup time in the cloud
A performance study on the vm startup time in the cloudA performance study on the vm startup time in the cloud
A performance study on the vm startup time in the cloud
 
Auto-Scaling to Minimize Cost and Meet Application Deadlines in Cloud Workflows
Auto-Scaling to Minimize Cost and Meet Application Deadlines in Cloud WorkflowsAuto-Scaling to Minimize Cost and Meet Application Deadlines in Cloud Workflows
Auto-Scaling to Minimize Cost and Meet Application Deadlines in Cloud Workflows
 
Ch5 - Project Management
Ch5 - Project ManagementCh5 - Project Management
Ch5 - Project Management
 
Task Scheduling and Asynchronous Processing Evolved. Zend Server Job Queue
Task Scheduling and Asynchronous Processing Evolved. Zend Server Job QueueTask Scheduling and Asynchronous Processing Evolved. Zend Server Job Queue
Task Scheduling and Asynchronous Processing Evolved. Zend Server Job Queue
 
Differentiating Algorithms of Cloud Task Scheduling Based on various Parameters
Differentiating Algorithms of Cloud Task Scheduling Based on various ParametersDifferentiating Algorithms of Cloud Task Scheduling Based on various Parameters
Differentiating Algorithms of Cloud Task Scheduling Based on various Parameters
 
An Introduction To Applied Evolutionary Meta Heuristics
An Introduction To Applied Evolutionary Meta HeuristicsAn Introduction To Applied Evolutionary Meta Heuristics
An Introduction To Applied Evolutionary Meta Heuristics
 

Similar a Scaling and scheduling to maximize application performance within budget constraints

Cost-Based Task Scheduling in Cloud Computing
Cost-Based Task Scheduling in Cloud ComputingCost-Based Task Scheduling in Cloud Computing
Cost-Based Task Scheduling in Cloud ComputingIRJET Journal
 
International Journal of Engineering Research and Development
International Journal of Engineering Research and DevelopmentInternational Journal of Engineering Research and Development
International Journal of Engineering Research and DevelopmentIJERD Editor
 
A HYPER-HEURISTIC METHOD FOR SCHEDULING THEJOBS IN CLOUD ENVIRONMENT
A HYPER-HEURISTIC METHOD FOR SCHEDULING THEJOBS IN CLOUD ENVIRONMENTA HYPER-HEURISTIC METHOD FOR SCHEDULING THEJOBS IN CLOUD ENVIRONMENT
A HYPER-HEURISTIC METHOD FOR SCHEDULING THEJOBS IN CLOUD ENVIRONMENTieijjournal
 
A HYPER-HEURISTIC METHOD FOR SCHEDULING THEJOBS IN CLOUD ENVIRONMENT
A HYPER-HEURISTIC METHOD FOR SCHEDULING THEJOBS IN CLOUD ENVIRONMENTA HYPER-HEURISTIC METHOD FOR SCHEDULING THEJOBS IN CLOUD ENVIRONMENT
A HYPER-HEURISTIC METHOD FOR SCHEDULING THEJOBS IN CLOUD ENVIRONMENTieijjournal1
 
Resource-Allocation.ppt
Resource-Allocation.pptResource-Allocation.ppt
Resource-Allocation.pptTecnicoItca
 
Simulation Based Workflow Scheduling for Scientific Application
Simulation Based Workflow Scheduling for Scientific ApplicationSimulation Based Workflow Scheduling for Scientific Application
Simulation Based Workflow Scheduling for Scientific ApplicationIJCSIS Research Publications
 
Deadline and Suffrage Aware Task Scheduling Approach for Cloud Environment
Deadline and Suffrage Aware Task Scheduling Approach for Cloud EnvironmentDeadline and Suffrage Aware Task Scheduling Approach for Cloud Environment
Deadline and Suffrage Aware Task Scheduling Approach for Cloud EnvironmentIRJET Journal
 
High Dimensionality Structures Selection for Efficient Economic Big data usin...
High Dimensionality Structures Selection for Efficient Economic Big data usin...High Dimensionality Structures Selection for Efficient Economic Big data usin...
High Dimensionality Structures Selection for Efficient Economic Big data usin...IRJET Journal
 
IRJET- Time and Resource Efficient Task Scheduling in Cloud Computing Environ...
IRJET- Time and Resource Efficient Task Scheduling in Cloud Computing Environ...IRJET- Time and Resource Efficient Task Scheduling in Cloud Computing Environ...
IRJET- Time and Resource Efficient Task Scheduling in Cloud Computing Environ...IRJET Journal
 
An Enhanced Throttled Load Balancing Approach for Cloud Environment
An Enhanced Throttled Load Balancing Approach for Cloud EnvironmentAn Enhanced Throttled Load Balancing Approach for Cloud Environment
An Enhanced Throttled Load Balancing Approach for Cloud EnvironmentIRJET Journal
 
LOAD BALANCING ALGORITHM TO IMPROVE RESPONSE TIME ON CLOUD COMPUTING
LOAD BALANCING ALGORITHM TO IMPROVE RESPONSE TIME ON CLOUD COMPUTINGLOAD BALANCING ALGORITHM TO IMPROVE RESPONSE TIME ON CLOUD COMPUTING
LOAD BALANCING ALGORITHM TO IMPROVE RESPONSE TIME ON CLOUD COMPUTINGijccsa
 
Load Balancing Algorithm to Improve Response Time on Cloud Computing
Load Balancing Algorithm to Improve Response Time on Cloud ComputingLoad Balancing Algorithm to Improve Response Time on Cloud Computing
Load Balancing Algorithm to Improve Response Time on Cloud Computingneirew J
 
Application of selective algorithm for effective resource provisioning in clo...
Application of selective algorithm for effective resource provisioning in clo...Application of selective algorithm for effective resource provisioning in clo...
Application of selective algorithm for effective resource provisioning in clo...ijccsa
 
A Survey on Service Request Scheduling in Cloud Based Architecture
A Survey on Service Request Scheduling in Cloud Based ArchitectureA Survey on Service Request Scheduling in Cloud Based Architecture
A Survey on Service Request Scheduling in Cloud Based ArchitectureIJSRD
 
A Survey on Service Request Scheduling in Cloud Based Architecture
A Survey on Service Request Scheduling in Cloud Based ArchitectureA Survey on Service Request Scheduling in Cloud Based Architecture
A Survey on Service Request Scheduling in Cloud Based ArchitectureIJSRD
 
Optimization of energy consumption in cloud computing datacenters
Optimization of energy consumption in cloud computing datacenters Optimization of energy consumption in cloud computing datacenters
Optimization of energy consumption in cloud computing datacenters IJECEIAES
 
Intelligent Workload Management in Virtualized Cloud Environment
Intelligent Workload Management in Virtualized Cloud EnvironmentIntelligent Workload Management in Virtualized Cloud Environment
Intelligent Workload Management in Virtualized Cloud EnvironmentIJTET Journal
 

Similar a Scaling and scheduling to maximize application performance within budget constraints (20)

Cost-Based Task Scheduling in Cloud Computing
Cost-Based Task Scheduling in Cloud ComputingCost-Based Task Scheduling in Cloud Computing
Cost-Based Task Scheduling in Cloud Computing
 
ch09s.ppt
ch09s.pptch09s.ppt
ch09s.ppt
 
International Journal of Engineering Research and Development
International Journal of Engineering Research and DevelopmentInternational Journal of Engineering Research and Development
International Journal of Engineering Research and Development
 
A HYPER-HEURISTIC METHOD FOR SCHEDULING THEJOBS IN CLOUD ENVIRONMENT
A HYPER-HEURISTIC METHOD FOR SCHEDULING THEJOBS IN CLOUD ENVIRONMENTA HYPER-HEURISTIC METHOD FOR SCHEDULING THEJOBS IN CLOUD ENVIRONMENT
A HYPER-HEURISTIC METHOD FOR SCHEDULING THEJOBS IN CLOUD ENVIRONMENT
 
A HYPER-HEURISTIC METHOD FOR SCHEDULING THEJOBS IN CLOUD ENVIRONMENT
A HYPER-HEURISTIC METHOD FOR SCHEDULING THEJOBS IN CLOUD ENVIRONMENTA HYPER-HEURISTIC METHOD FOR SCHEDULING THEJOBS IN CLOUD ENVIRONMENT
A HYPER-HEURISTIC METHOD FOR SCHEDULING THEJOBS IN CLOUD ENVIRONMENT
 
Resource-Allocation.ppt
Resource-Allocation.pptResource-Allocation.ppt
Resource-Allocation.ppt
 
Simulation Based Workflow Scheduling for Scientific Application
Simulation Based Workflow Scheduling for Scientific ApplicationSimulation Based Workflow Scheduling for Scientific Application
Simulation Based Workflow Scheduling for Scientific Application
 
Deadline and Suffrage Aware Task Scheduling Approach for Cloud Environment
Deadline and Suffrage Aware Task Scheduling Approach for Cloud EnvironmentDeadline and Suffrage Aware Task Scheduling Approach for Cloud Environment
Deadline and Suffrage Aware Task Scheduling Approach for Cloud Environment
 
CE427-Chp6-Resource-Allocation-24Apr2019.pptx
CE427-Chp6-Resource-Allocation-24Apr2019.pptxCE427-Chp6-Resource-Allocation-24Apr2019.pptx
CE427-Chp6-Resource-Allocation-24Apr2019.pptx
 
High Dimensionality Structures Selection for Efficient Economic Big data usin...
High Dimensionality Structures Selection for Efficient Economic Big data usin...High Dimensionality Structures Selection for Efficient Economic Big data usin...
High Dimensionality Structures Selection for Efficient Economic Big data usin...
 
IRJET- Time and Resource Efficient Task Scheduling in Cloud Computing Environ...
IRJET- Time and Resource Efficient Task Scheduling in Cloud Computing Environ...IRJET- Time and Resource Efficient Task Scheduling in Cloud Computing Environ...
IRJET- Time and Resource Efficient Task Scheduling in Cloud Computing Environ...
 
An Enhanced Throttled Load Balancing Approach for Cloud Environment
An Enhanced Throttled Load Balancing Approach for Cloud EnvironmentAn Enhanced Throttled Load Balancing Approach for Cloud Environment
An Enhanced Throttled Load Balancing Approach for Cloud Environment
 
LOAD BALANCING ALGORITHM TO IMPROVE RESPONSE TIME ON CLOUD COMPUTING
LOAD BALANCING ALGORITHM TO IMPROVE RESPONSE TIME ON CLOUD COMPUTINGLOAD BALANCING ALGORITHM TO IMPROVE RESPONSE TIME ON CLOUD COMPUTING
LOAD BALANCING ALGORITHM TO IMPROVE RESPONSE TIME ON CLOUD COMPUTING
 
Load Balancing Algorithm to Improve Response Time on Cloud Computing
Load Balancing Algorithm to Improve Response Time on Cloud ComputingLoad Balancing Algorithm to Improve Response Time on Cloud Computing
Load Balancing Algorithm to Improve Response Time on Cloud Computing
 
Application of selective algorithm for effective resource provisioning in clo...
Application of selective algorithm for effective resource provisioning in clo...Application of selective algorithm for effective resource provisioning in clo...
Application of selective algorithm for effective resource provisioning in clo...
 
A Survey on Service Request Scheduling in Cloud Based Architecture
A Survey on Service Request Scheduling in Cloud Based ArchitectureA Survey on Service Request Scheduling in Cloud Based Architecture
A Survey on Service Request Scheduling in Cloud Based Architecture
 
A Survey on Service Request Scheduling in Cloud Based Architecture
A Survey on Service Request Scheduling in Cloud Based ArchitectureA Survey on Service Request Scheduling in Cloud Based Architecture
A Survey on Service Request Scheduling in Cloud Based Architecture
 
Optimization of energy consumption in cloud computing datacenters
Optimization of energy consumption in cloud computing datacenters Optimization of energy consumption in cloud computing datacenters
Optimization of energy consumption in cloud computing datacenters
 
-PM-09-7.ppt
-PM-09-7.ppt-PM-09-7.ppt
-PM-09-7.ppt
 
Intelligent Workload Management in Virtualized Cloud Environment
Intelligent Workload Management in Virtualized Cloud EnvironmentIntelligent Workload Management in Virtualized Cloud Environment
Intelligent Workload Management in Virtualized Cloud Environment
 

Último

Advanced Computer Architecture – An Introduction
Advanced Computer Architecture – An IntroductionAdvanced Computer Architecture – An Introduction
Advanced Computer Architecture – An IntroductionDilum Bandara
 
WordPress Websites for Engineers: Elevate Your Brand
WordPress Websites for Engineers: Elevate Your BrandWordPress Websites for Engineers: Elevate Your Brand
WordPress Websites for Engineers: Elevate Your Brandgvaughan
 
DevEX - reference for building teams, processes, and platforms
DevEX - reference for building teams, processes, and platformsDevEX - reference for building teams, processes, and platforms
DevEX - reference for building teams, processes, and platformsSergiu Bodiu
 
Are Multi-Cloud and Serverless Good or Bad?
Are Multi-Cloud and Serverless Good or Bad?Are Multi-Cloud and Serverless Good or Bad?
Are Multi-Cloud and Serverless Good or Bad?Mattias Andersson
 
Unraveling Multimodality with Large Language Models.pdf
Unraveling Multimodality with Large Language Models.pdfUnraveling Multimodality with Large Language Models.pdf
Unraveling Multimodality with Large Language Models.pdfAlex Barbosa Coqueiro
 
Unleash Your Potential - Namagunga Girls Coding Club
Unleash Your Potential - Namagunga Girls Coding ClubUnleash Your Potential - Namagunga Girls Coding Club
Unleash Your Potential - Namagunga Girls Coding ClubKalema Edgar
 
SIP trunking in Janus @ Kamailio World 2024
SIP trunking in Janus @ Kamailio World 2024SIP trunking in Janus @ Kamailio World 2024
SIP trunking in Janus @ Kamailio World 2024Lorenzo Miniero
 
Hyperautomation and AI/ML: A Strategy for Digital Transformation Success.pdf
Hyperautomation and AI/ML: A Strategy for Digital Transformation Success.pdfHyperautomation and AI/ML: A Strategy for Digital Transformation Success.pdf
Hyperautomation and AI/ML: A Strategy for Digital Transformation Success.pdfPrecisely
 
Developer Data Modeling Mistakes: From Postgres to NoSQL
Developer Data Modeling Mistakes: From Postgres to NoSQLDeveloper Data Modeling Mistakes: From Postgres to NoSQL
Developer Data Modeling Mistakes: From Postgres to NoSQLScyllaDB
 
"ML in Production",Oleksandr Bagan
"ML in Production",Oleksandr Bagan"ML in Production",Oleksandr Bagan
"ML in Production",Oleksandr BaganFwdays
 
DevoxxFR 2024 Reproducible Builds with Apache Maven
DevoxxFR 2024 Reproducible Builds with Apache MavenDevoxxFR 2024 Reproducible Builds with Apache Maven
DevoxxFR 2024 Reproducible Builds with Apache MavenHervé Boutemy
 
The State of Passkeys with FIDO Alliance.pptx
The State of Passkeys with FIDO Alliance.pptxThe State of Passkeys with FIDO Alliance.pptx
The State of Passkeys with FIDO Alliance.pptxLoriGlavin3
 
Ensuring Technical Readiness For Copilot in Microsoft 365
Ensuring Technical Readiness For Copilot in Microsoft 365Ensuring Technical Readiness For Copilot in Microsoft 365
Ensuring Technical Readiness For Copilot in Microsoft 3652toLead Limited
 
Tampa BSides - Chef's Tour of Microsoft Security Adoption Framework (SAF)
Tampa BSides - Chef's Tour of Microsoft Security Adoption Framework (SAF)Tampa BSides - Chef's Tour of Microsoft Security Adoption Framework (SAF)
Tampa BSides - Chef's Tour of Microsoft Security Adoption Framework (SAF)Mark Simos
 
DSPy a system for AI to Write Prompts and Do Fine Tuning
DSPy a system for AI to Write Prompts and Do Fine TuningDSPy a system for AI to Write Prompts and Do Fine Tuning
DSPy a system for AI to Write Prompts and Do Fine TuningLars Bell
 
New from BookNet Canada for 2024: Loan Stars - Tech Forum 2024
New from BookNet Canada for 2024: Loan Stars - Tech Forum 2024New from BookNet Canada for 2024: Loan Stars - Tech Forum 2024
New from BookNet Canada for 2024: Loan Stars - Tech Forum 2024BookNet Canada
 
Connect Wave/ connectwave Pitch Deck Presentation
Connect Wave/ connectwave Pitch Deck PresentationConnect Wave/ connectwave Pitch Deck Presentation
Connect Wave/ connectwave Pitch Deck PresentationSlibray Presentation
 
Commit 2024 - Secret Management made easy
Commit 2024 - Secret Management made easyCommit 2024 - Secret Management made easy
Commit 2024 - Secret Management made easyAlfredo García Lavilla
 
How AI, OpenAI, and ChatGPT impact business and software.
How AI, OpenAI, and ChatGPT impact business and software.How AI, OpenAI, and ChatGPT impact business and software.
How AI, OpenAI, and ChatGPT impact business and software.Curtis Poe
 
Moving Beyond Passwords: FIDO Paris Seminar.pdf
Moving Beyond Passwords: FIDO Paris Seminar.pdfMoving Beyond Passwords: FIDO Paris Seminar.pdf
Moving Beyond Passwords: FIDO Paris Seminar.pdfLoriGlavin3
 

Último (20)

Advanced Computer Architecture – An Introduction
Advanced Computer Architecture – An IntroductionAdvanced Computer Architecture – An Introduction
Advanced Computer Architecture – An Introduction
 
WordPress Websites for Engineers: Elevate Your Brand
WordPress Websites for Engineers: Elevate Your BrandWordPress Websites for Engineers: Elevate Your Brand
WordPress Websites for Engineers: Elevate Your Brand
 
DevEX - reference for building teams, processes, and platforms
DevEX - reference for building teams, processes, and platformsDevEX - reference for building teams, processes, and platforms
DevEX - reference for building teams, processes, and platforms
 
Are Multi-Cloud and Serverless Good or Bad?
Are Multi-Cloud and Serverless Good or Bad?Are Multi-Cloud and Serverless Good or Bad?
Are Multi-Cloud and Serverless Good or Bad?
 
Unraveling Multimodality with Large Language Models.pdf
Unraveling Multimodality with Large Language Models.pdfUnraveling Multimodality with Large Language Models.pdf
Unraveling Multimodality with Large Language Models.pdf
 
Unleash Your Potential - Namagunga Girls Coding Club
Unleash Your Potential - Namagunga Girls Coding ClubUnleash Your Potential - Namagunga Girls Coding Club
Unleash Your Potential - Namagunga Girls Coding Club
 
SIP trunking in Janus @ Kamailio World 2024
SIP trunking in Janus @ Kamailio World 2024SIP trunking in Janus @ Kamailio World 2024
SIP trunking in Janus @ Kamailio World 2024
 
Hyperautomation and AI/ML: A Strategy for Digital Transformation Success.pdf
Hyperautomation and AI/ML: A Strategy for Digital Transformation Success.pdfHyperautomation and AI/ML: A Strategy for Digital Transformation Success.pdf
Hyperautomation and AI/ML: A Strategy for Digital Transformation Success.pdf
 
Developer Data Modeling Mistakes: From Postgres to NoSQL
Developer Data Modeling Mistakes: From Postgres to NoSQLDeveloper Data Modeling Mistakes: From Postgres to NoSQL
Developer Data Modeling Mistakes: From Postgres to NoSQL
 
"ML in Production",Oleksandr Bagan
"ML in Production",Oleksandr Bagan"ML in Production",Oleksandr Bagan
"ML in Production",Oleksandr Bagan
 
DevoxxFR 2024 Reproducible Builds with Apache Maven
DevoxxFR 2024 Reproducible Builds with Apache MavenDevoxxFR 2024 Reproducible Builds with Apache Maven
DevoxxFR 2024 Reproducible Builds with Apache Maven
 
The State of Passkeys with FIDO Alliance.pptx
The State of Passkeys with FIDO Alliance.pptxThe State of Passkeys with FIDO Alliance.pptx
The State of Passkeys with FIDO Alliance.pptx
 
Ensuring Technical Readiness For Copilot in Microsoft 365
Ensuring Technical Readiness For Copilot in Microsoft 365Ensuring Technical Readiness For Copilot in Microsoft 365
Ensuring Technical Readiness For Copilot in Microsoft 365
 
Tampa BSides - Chef's Tour of Microsoft Security Adoption Framework (SAF)
Tampa BSides - Chef's Tour of Microsoft Security Adoption Framework (SAF)Tampa BSides - Chef's Tour of Microsoft Security Adoption Framework (SAF)
Tampa BSides - Chef's Tour of Microsoft Security Adoption Framework (SAF)
 
DSPy a system for AI to Write Prompts and Do Fine Tuning
DSPy a system for AI to Write Prompts and Do Fine TuningDSPy a system for AI to Write Prompts and Do Fine Tuning
DSPy a system for AI to Write Prompts and Do Fine Tuning
 
New from BookNet Canada for 2024: Loan Stars - Tech Forum 2024
New from BookNet Canada for 2024: Loan Stars - Tech Forum 2024New from BookNet Canada for 2024: Loan Stars - Tech Forum 2024
New from BookNet Canada for 2024: Loan Stars - Tech Forum 2024
 
Connect Wave/ connectwave Pitch Deck Presentation
Connect Wave/ connectwave Pitch Deck PresentationConnect Wave/ connectwave Pitch Deck Presentation
Connect Wave/ connectwave Pitch Deck Presentation
 
Commit 2024 - Secret Management made easy
Commit 2024 - Secret Management made easyCommit 2024 - Secret Management made easy
Commit 2024 - Secret Management made easy
 
How AI, OpenAI, and ChatGPT impact business and software.
How AI, OpenAI, and ChatGPT impact business and software.How AI, OpenAI, and ChatGPT impact business and software.
How AI, OpenAI, and ChatGPT impact business and software.
 
Moving Beyond Passwords: FIDO Paris Seminar.pdf
Moving Beyond Passwords: FIDO Paris Seminar.pdfMoving Beyond Passwords: FIDO Paris Seminar.pdf
Moving Beyond Passwords: FIDO Paris Seminar.pdf
 

Scaling and scheduling to maximize application performance within budget constraints

  • 1. Ming Mao, Marty Humphrey CS Department, UVa Scaling and Scheduling to Maximize Application Performance within Budget Constraints in Cloud Workflows IPDPS 2013 (May 21st 2013) 1
  • 2. 2  Dynamic scalability and cost saving are two of the most important factors when considering cloud adoption Two major benefits - dynamic scalability and cost A survey from 39 major technology companies [1]  Cloud benefits  On-demand self-services  Broad network access  Resource pooling  Rapid elasticity  Measured services  Cheaper maintenance  …… Why do you move into the cloud?
  • 3. 3  Dynamic scalability – the ability to acquire/release resources in response to demand dynamically  Dynamic scalability challenge → It relies on the users to tell the size of resource pool  Over-provisioning → cost more than necessary, offset cloud advantages  Under-provisioning → hurt application performance, cannot meet service level agreements and lose application customers Cloud dynamic scalability over-provisioning under-provisioning
  • 4. 4  Problem - What resources should be acquired/released in the cloud, and how should the computing activities be mapped to the cloud resources, so that the application performance can be maximized within the budget constrains?  In this paper, we discuss limited budget case  The unlimited budget case was discussed in our SC 11 paper  Solution - This paper argues that an automatic resource provisioning and allocation mechanism, i.e., an auto-scaling solution – is the key to successful cloud adoption. Essentially, an auto-scaling solution needs to answer the following two questions:  Capacity determination (or resource provisioning)  what types of resources, how much and for how long  Job scheduling (or resource allocation)  map computing activities onto the cloud resources Problem statement
  • 5. 5  An application consists of service components. A workflow goes through different service components and therefore consists of multiple connected tasks  Workload is a stream of workflow jobs not known in advance  Task precedence constraints need to be preserved  Jobs have individual priorities Service oriented architecture (SOA) & workflow jobs
  • 6. 6 Minimize job turnaround time within budget constraints Problem formulation  Problem terminology  Cloud application  app = {Si}  Job class  J = {DAG(Si), priorityJ| Si ∈ app}  Cloud VM  VMv = {[𝐽 𝑆 𝑖]v , cv , lagv}  Workload  Wt = 𝑗𝑜𝑏𝐽 𝑆 𝑖 𝑗𝑜𝑏𝐽𝑆 𝑖  Scaling plan  Scalingt = {VMv → Nv}  Scheduling plan  Schedulet = { 𝑗𝐽 𝑆 𝑖 →VMv}  Goal  Min( 𝑗𝑜𝑏𝑡𝑢𝑟𝑛𝑎𝑟𝑜𝑢𝑛𝑑 × 𝑝𝑟𝑖𝑜𝑟𝑖𝑡𝑦/𝑗𝑜𝑏 𝑝𝑟𝑖𝑜𝑟𝑖𝑡𝑦𝑗𝑜𝑏 ) &&  Cost(app) <= B (budget, dollars/hour)  Target - The service provider has a limited budget and aims to maximize the application performance.  Solution idea – a monitor-control loop that makes scaling and scheduling decisions based on updated workload and VM information
  • 7. 7  Scheduling-first  Idea – allocate application budget to individual jobs based on priorities and schedule tasks within job budget  Step 1 – Distribute budget: 𝐵𝑗 = 𝐵 × 𝑝𝑗/ 𝑝𝑗𝑗  Step 2 – Schedule tasks  for each job, schedule as many tasks as possible on their fast machines  Step 3 – Consolidate budget  return job budget to the application  the application uses the remaining budget collected from individual jobs to schedule high priority tasks  Step 4 – Acquire instance  acquire instances and execute tasks based on the determined schedule plans Minimize job turnaround time within budget constraints Solution: scheduling-first
  • 8. 8  Scheduling-first  Step 1 – Distribute budget: 𝐵𝑗 = 𝐵 × 𝑝𝑗/ 𝑝𝑗𝑗 Minimize job turnaround time within budget constraints Solution: scheduling-first  Step 2 – Schedule tasks e.g. Budget(B) = $1/h; Large(L) = $0.5/h; Medium(M) = $0.3/h; Small(S) = $0.1/h;  Step 1: job1 and job2 have the same priority, job1 → $0.5/h, job2 → $0.5/h  Step 2: job1(T1) → $0.5(L); job2(T5) → $0.5(L);  Step 3: job1(T2+T3) → $0.5(S+M); job2(T6) → $0.5(L); job1 returns $0.1 to system; job2(T7) → $0.1(S);  Step 4 acquire instances when necessary  Step 3 – Consolidate budget  Step 4 – Acquire instance
  • 9. 9 Minimize job turnaround time within budget constraints Solution: scaling-first  Scaling-first  Idea – determine the computing capacity by looking at the overall workload and schedule tasks based on priority  Step 1 – determine the VMs  assume tasks run on their fastest machines and calculate the cost Cfast for the next hour  acquire VMs proportionally based on Budget/Cfast  Step 2 – consolidate budget  use the remaining the budget to purchase new machines.  Step 3 – schedule tasks  schedule tasks based on task priority
  • 10. 10 Minimize job turnaround time within budget constraints Solution: scaling-first  Scaling-first  Step 1 – determine the VMs  Step 2 – consolidate budget  Step 3 – schedule tasks  Step 1: assume tasks run on fastest machines and calculate Cfast and acquire VMs proportionally based on B/Cfast,  Step 2: the remaining $0.5 can be used to purchase 1 L machine  Step 3: tasks are scheduled based on their priorities
  • 11. 11  Instance consolidation  Schedule tasks on different VM types to save partial instance hour cost  Budget allocation schemes  Evenly distributed – e.g. daily x/365, hourly x/8760  Based on workload – e.g. high on busy times, low on non-busy times  Workload prediction – $/hour → $/job Minimize job turnaround time within budget constraints Other considerations
  • 12.  Workload patterns  Application models 12  Time  72 hours  Task execution  Randomly generated  VM lag  5 min Minimize job turnaround time within budget constraints Evaluation – experiment setup  Baseline  Standard VM Type Price Micro $0.02/hour Standard $0.080/hour High-CPU $0.66/hour High-Memory $0.45/hour Extra-Large $1.3/hour
  • 13. 13 Minimize job turnaround time within budget constraints Evaluation – job turnaround time  above – weighted average job turnaround time for the hybrid application and cycle workload pattern  Scheduling-first and scaling-first can save 9.8%- 45.2% cost compared to the standard machine choice.  Scaling-first works better under small budget ranges while scheduling-first works better under large budget ranges.
  • 14. 14 Minimize job turnaround time within budget constraints Evaluation – sensitivity to inaccurate parameters  left – scheduling-first’s sensitivity to inaccurate parameters (Hybrid application + Cycle workload pattern)  right – scaling-first’s sensitivity to inaccurate parameters (Hybrid application + Cycle workload pattern)  When the estimation error is within ±20%, the job turnaround time shows -10.2% – 16.7% difference.  When the task estimation error reaches ±60%, the performance of both algorithms shows significant degradation (more than ±25% difference)
  • 15. 15 Minimize job turnaround time within budget constraints Evaluation – instance consolidation  left – job turnaround time / resource utilization of scheduling-first’s instance consolidation (Hybrid application + Cycle workload pattern)  right – job turnaround time / resource utilization of scaling-first’s instance consolidation (Hybrid application + Cycle workload pattern)  When budget is low or high, the improvement is small. When the budget is in between, the improvement is more significant (e.g. utilization rate improves 2.2% to 19.9% when the budget is between $15/hour and $25/hour).  Scaling-first benefits more from instance consolidation process than scheduling-first
  • 16. 16  Conclusions  choose appropriate VM types based on the workload.  Scheduling-first and scaling-first are trade-offs between the task execution time and waiting time.  As long as the VM performance can be correctly ranked, the proposed mechanisms have good tolerance to inaccurate parameters.  Instance consolidation is an efficient strategy to save partial instance hours and improve resource utilization.  Future work  Other billing models – reserved instances, spot instances, $/min  Maximize application performance within budget constraints for data-intensive applications  Hybrid and federate cloud environments  Develop evaluation benchmarks and simulation platforms Conclusion and future work