SlideShare une entreprise Scribd logo
1  sur  11
Télécharger pour lire hors ligne
Chris Condo
ccondo@outlook.com
August 20, 2014
Azure & HPC Configuration
HPC = High Performance Computing
• HPC is Microsoft’s “Big Compute” platform
• Parallel and batch compute capabilities.
• Applications:
• Actuarial sciences.
• Financial Risk Modeling
• Digital content creation, transcoding.
• Genetics research.
• Head Node hosts a control panel which allows you
to create cluster configurations.
• The head node sends work the compute node
clusters.
• The work is split up into work units and executed in
parallel.
Typical on premise HPC configuration
HEAD NODE
ON PREMISE COMPUTE NODES
Hybrid HPC-Azure configuration
HEAD NODE
ON PREM COMPUTE NODES
AZURE COMPUTE NODES
• Add capacity with worker nodes in the cloud.
• Useful for scenarios such as seasonal bursts
where extra capacity is needed occasionally.
• Its not necessarily about speed, its about
throughput.
Azure Compute Node Size
• Use either A8 or A9 class machines.
• Problem sets can be configured to
be processed in parallel across 8 or
16 cores.
• These machines come at an added
cost.
• Managing that resource is
important to controlling $$$ spent.
Cluster Manager running on the head node
Cluster Manager starting Azure Nodes.
• There’s manual steps
needed to start/stop
jobs.
• These steps can be
automated via
PowerShell
Powershell can replace manual interaction.
• Your application can be written to execute Powershell to add
Azure capacity as needed.
• Add nodes to your HPC Cluster dynamically
• Get-HpcNodeTemplate $HpcNodeTemplate -Scheduler $Scheduler | Add-
HpcNodeSet -Quantity $Quantity -Size $Size -Scheduler $Scheduler
• Start or Stop the nodes
• Get-HpcNodeTemplate $HpcNodeTemplate -Scheduler $Scheduler | Start-
HpcNodeSet -Scheduler $Scheduler
• Shut down the nodes when the job is done, don’t waste $$$
with a VM that’s not doing work.
Windows Azure Management Libraries
• A Natural .NET Framework that maps cleanly to the underlying REST API
• Introduced late 2013
• Supports the Portable Class Library (PCL).
• Ships as a NuGet Package you install from VS
• Designed to use C# 5 async tasks
• Easy to code with C#
using Microsoft.WindowsAzure;
using Microsoft.WindowsAzure.Management.Compute;
using Microsoft.WindowsAzure.Management.Storage;
Do cool stuff like create Azure VMs from a desktop.
• Deploy cloud services from your .Net applications
• Eg: Create a VM
internal async Task CreateVM()
{ await _computeManagementClient.VirtualMachines.CreateAsync(
“MyAzureService",
“MyAzureDeployment",
new VirtualMachineCreateParameters
{ RoleName = “MyName",
RoleSize = VirtualMachineRoleSize.ExtraSmall
},
new System.Threading.CancellationToken());
}
Summary
• HPC is specialized for certain industries that require massive computing
power and can make use of parallel processing.
• Azure “Burst to cloud” is an effective means to add capacity.
• Automating the starting and stopping of your Azure resources can help
control costs.
• Powershell and/or WAML are 2 ways to develop operational scripts to
automate Hybrid-HPC work management.

Contenu connexe

Tendances

High performance computing
High performance computingHigh performance computing
High performance computing
Guy Tel-Zur
 
Timely Year Two: Lessons Learned Building a Scalable Metrics Analytic System
Timely Year Two: Lessons Learned Building a Scalable Metrics Analytic SystemTimely Year Two: Lessons Learned Building a Scalable Metrics Analytic System
Timely Year Two: Lessons Learned Building a Scalable Metrics Analytic System
Accumulo Summit
 
Scaling MLOps on NVIDIA DGX Systems
Scaling MLOps on NVIDIA DGX SystemsScaling MLOps on NVIDIA DGX Systems
Scaling MLOps on NVIDIA DGX Systems
cnvrg.io AI OS - Hands-on ML Workshops
 
Load Balancing In Cloud Computing newppt
Load Balancing In Cloud Computing newpptLoad Balancing In Cloud Computing newppt
Load Balancing In Cloud Computing newppt
Utshab Saha
 

Tendances (20)

Dr. Ike Nassi, Founder, TidalScale at MLconf NYC - 4/15/16
Dr. Ike Nassi, Founder, TidalScale at MLconf NYC - 4/15/16Dr. Ike Nassi, Founder, TidalScale at MLconf NYC - 4/15/16
Dr. Ike Nassi, Founder, TidalScale at MLconf NYC - 4/15/16
 
High performance computing
High performance computingHigh performance computing
High performance computing
 
Which Is Deeper - Comparison Of Deep Learning Frameworks On Spark
 Which Is Deeper - Comparison Of Deep Learning Frameworks On Spark Which Is Deeper - Comparison Of Deep Learning Frameworks On Spark
Which Is Deeper - Comparison Of Deep Learning Frameworks On Spark
 
Modest scale HPC on Azure using CGYRO
Modest scale HPC on Azure using CGYROModest scale HPC on Azure using CGYRO
Modest scale HPC on Azure using CGYRO
 
AWS Webcast - An Introduction to High Performance Computing on AWS
AWS Webcast - An Introduction to High Performance Computing on AWSAWS Webcast - An Introduction to High Performance Computing on AWS
AWS Webcast - An Introduction to High Performance Computing on AWS
 
Timely Year Two: Lessons Learned Building a Scalable Metrics Analytic System
Timely Year Two: Lessons Learned Building a Scalable Metrics Analytic SystemTimely Year Two: Lessons Learned Building a Scalable Metrics Analytic System
Timely Year Two: Lessons Learned Building a Scalable Metrics Analytic System
 
Accelerated Machine Learning with RAPIDS and MLflow, Nvidia/RAPIDS
Accelerated Machine Learning with RAPIDS and MLflow, Nvidia/RAPIDSAccelerated Machine Learning with RAPIDS and MLflow, Nvidia/RAPIDS
Accelerated Machine Learning with RAPIDS and MLflow, Nvidia/RAPIDS
 
Scientific Computing With Amazon Web Services
Scientific Computing With Amazon Web ServicesScientific Computing With Amazon Web Services
Scientific Computing With Amazon Web Services
 
High Performance Computing: an Introduction for the Society of Actuaries
High Performance Computing: an Introduction for the Society of ActuariesHigh Performance Computing: an Introduction for the Society of Actuaries
High Performance Computing: an Introduction for the Society of Actuaries
 
Scaling MLOps on NVIDIA DGX Systems
Scaling MLOps on NVIDIA DGX SystemsScaling MLOps on NVIDIA DGX Systems
Scaling MLOps on NVIDIA DGX Systems
 
OpenNebula TechDay Boston 2015 - Future of Information Storage with ISS Super...
OpenNebula TechDay Boston 2015 - Future of Information Storage with ISS Super...OpenNebula TechDay Boston 2015 - Future of Information Storage with ISS Super...
OpenNebula TechDay Boston 2015 - Future of Information Storage with ISS Super...
 
12.07.2017 Docker Meetup - POSTGRE SQL ON KUBERNETES
12.07.2017 Docker Meetup - POSTGRE SQL ON KUBERNETES12.07.2017 Docker Meetup - POSTGRE SQL ON KUBERNETES
12.07.2017 Docker Meetup - POSTGRE SQL ON KUBERNETES
 
Nyc kubernetes Meetup - Kubeflow Lightning talk
Nyc kubernetes Meetup - Kubeflow Lightning talkNyc kubernetes Meetup - Kubeflow Lightning talk
Nyc kubernetes Meetup - Kubeflow Lightning talk
 
Query Anything, Anywhere with Kubernetes
Query Anything, Anywhere with KubernetesQuery Anything, Anywhere with Kubernetes
Query Anything, Anywhere with Kubernetes
 
Prediction as a service with ensemble model in SparkML and Python ScikitLearn
Prediction as a service with ensemble model in SparkML and Python ScikitLearnPrediction as a service with ensemble model in SparkML and Python ScikitLearn
Prediction as a service with ensemble model in SparkML and Python ScikitLearn
 
HPC in the Cloud
HPC in the CloudHPC in the Cloud
HPC in the Cloud
 
Surge 2013: Maximizing Scalability, Resiliency, and Engineering Velocity in t...
Surge 2013: Maximizing Scalability, Resiliency, and Engineering Velocity in t...Surge 2013: Maximizing Scalability, Resiliency, and Engineering Velocity in t...
Surge 2013: Maximizing Scalability, Resiliency, and Engineering Velocity in t...
 
Dask for Analytics
Dask for AnalyticsDask for Analytics
Dask for Analytics
 
Advanced Spark and TensorFlow Meetup - Dec 12 2017 - Dong Meng, MapR + Kubern...
Advanced Spark and TensorFlow Meetup - Dec 12 2017 - Dong Meng, MapR + Kubern...Advanced Spark and TensorFlow Meetup - Dec 12 2017 - Dong Meng, MapR + Kubern...
Advanced Spark and TensorFlow Meetup - Dec 12 2017 - Dong Meng, MapR + Kubern...
 
Load Balancing In Cloud Computing newppt
Load Balancing In Cloud Computing newpptLoad Balancing In Cloud Computing newppt
Load Balancing In Cloud Computing newppt
 

En vedette

En vedette (7)

Paz
PazPaz
Paz
 
Cr7
Cr7Cr7
Cr7
 
Final year project center coimbatore
Final year project center coimbatoreFinal year project center coimbatore
Final year project center coimbatore
 
Giấy Can - Văn phòng phẩm 365
Giấy Can - Văn phòng phẩm 365Giấy Can - Văn phòng phẩm 365
Giấy Can - Văn phòng phẩm 365
 
SITEC Ec Class - Proses E- Runcit by Coach Fatin Fadila
SITEC Ec Class - Proses E- Runcit by Coach Fatin FadilaSITEC Ec Class - Proses E- Runcit by Coach Fatin Fadila
SITEC Ec Class - Proses E- Runcit by Coach Fatin Fadila
 
Vektor (pertemuan 1)
Vektor (pertemuan 1)Vektor (pertemuan 1)
Vektor (pertemuan 1)
 
Libro Codelpa - WEB
Libro Codelpa - WEBLibro Codelpa - WEB
Libro Codelpa - WEB
 

Similaire à HybridAzureCloud

Apache Druid Auto Scale-out/in for Streaming Data Ingestion on Kubernetes
Apache Druid Auto Scale-out/in for Streaming Data Ingestion on KubernetesApache Druid Auto Scale-out/in for Streaming Data Ingestion on Kubernetes
Apache Druid Auto Scale-out/in for Streaming Data Ingestion on Kubernetes
DataWorks Summit
 

Similaire à HybridAzureCloud (20)

HPC in the Cloud
HPC in the CloudHPC in the Cloud
HPC in the Cloud
 
Deep Dive: OpenStack Summit (Red Hat Summit 2014)
Deep Dive: OpenStack Summit (Red Hat Summit 2014)Deep Dive: OpenStack Summit (Red Hat Summit 2014)
Deep Dive: OpenStack Summit (Red Hat Summit 2014)
 
Building a Just-in-Time Application Stack for Analysts
Building a Just-in-Time Application Stack for AnalystsBuilding a Just-in-Time Application Stack for Analysts
Building a Just-in-Time Application Stack for Analysts
 
SCaLE 20X: Kubernetes Cloud Cost Monitoring with OpenCost & Optimization Stra...
SCaLE 20X: Kubernetes Cloud Cost Monitoring with OpenCost & Optimization Stra...SCaLE 20X: Kubernetes Cloud Cost Monitoring with OpenCost & Optimization Stra...
SCaLE 20X: Kubernetes Cloud Cost Monitoring with OpenCost & Optimization Stra...
 
High Performance Computing Pitch Deck
High Performance Computing Pitch DeckHigh Performance Computing Pitch Deck
High Performance Computing Pitch Deck
 
Distributed Tensorflow with Kubernetes - data2day - Jakob Karalus
Distributed Tensorflow with Kubernetes - data2day - Jakob KaralusDistributed Tensorflow with Kubernetes - data2day - Jakob Karalus
Distributed Tensorflow with Kubernetes - data2day - Jakob Karalus
 
Building HPC Clusters as Code in the (Almost) Infinite Cloud | AWS Public Sec...
Building HPC Clusters as Code in the (Almost) Infinite Cloud | AWS Public Sec...Building HPC Clusters as Code in the (Almost) Infinite Cloud | AWS Public Sec...
Building HPC Clusters as Code in the (Almost) Infinite Cloud | AWS Public Sec...
 
Migrating existing open source machine learning to azure
Migrating existing open source machine learning to azureMigrating existing open source machine learning to azure
Migrating existing open source machine learning to azure
 
Cloud Bursting 101: What to do When Cloud Computing Demand Exceeds Capacity
Cloud Bursting 101: What to do When Cloud Computing Demand Exceeds CapacityCloud Bursting 101: What to do When Cloud Computing Demand Exceeds Capacity
Cloud Bursting 101: What to do When Cloud Computing Demand Exceeds Capacity
 
The Convergence of HPC and Deep Learning
The Convergence of HPC and Deep LearningThe Convergence of HPC and Deep Learning
The Convergence of HPC and Deep Learning
 
Cloud-Native DevOps: Simplifying application lifecycle management with AWS | ...
Cloud-Native DevOps: Simplifying application lifecycle management with AWS | ...Cloud-Native DevOps: Simplifying application lifecycle management with AWS | ...
Cloud-Native DevOps: Simplifying application lifecycle management with AWS | ...
 
AWS ECS workshop
AWS ECS workshopAWS ECS workshop
AWS ECS workshop
 
9th docker meetup 2016.07.13
9th docker meetup 2016.07.139th docker meetup 2016.07.13
9th docker meetup 2016.07.13
 
Speed up R with parallel programming in the Cloud
Speed up R with parallel programming in the CloudSpeed up R with parallel programming in the Cloud
Speed up R with parallel programming in the Cloud
 
Cassandra Summit 2014: Apache Cassandra Best Practices at Ebay
Cassandra Summit 2014: Apache Cassandra Best Practices at EbayCassandra Summit 2014: Apache Cassandra Best Practices at Ebay
Cassandra Summit 2014: Apache Cassandra Best Practices at Ebay
 
Apache Druid Auto Scale-out/in for Streaming Data Ingestion on Kubernetes
Apache Druid Auto Scale-out/in for Streaming Data Ingestion on KubernetesApache Druid Auto Scale-out/in for Streaming Data Ingestion on Kubernetes
Apache Druid Auto Scale-out/in for Streaming Data Ingestion on Kubernetes
 
AWS re:Invent 2016: Building HPC Clusters as Code in the (Almost) Infinite Cl...
AWS re:Invent 2016: Building HPC Clusters as Code in the (Almost) Infinite Cl...AWS re:Invent 2016: Building HPC Clusters as Code in the (Almost) Infinite Cl...
AWS re:Invent 2016: Building HPC Clusters as Code in the (Almost) Infinite Cl...
 
Infrastructure as Code
Infrastructure as CodeInfrastructure as Code
Infrastructure as Code
 
Big Data Day LA 2016/ Big Data Track - How To Use Impala and Kudu To Optimize...
Big Data Day LA 2016/ Big Data Track - How To Use Impala and Kudu To Optimize...Big Data Day LA 2016/ Big Data Track - How To Use Impala and Kudu To Optimize...
Big Data Day LA 2016/ Big Data Track - How To Use Impala and Kudu To Optimize...
 
Sandstone HPC: A Domain General Gateway for New Users
Sandstone HPC: A Domain General Gateway for New UsersSandstone HPC: A Domain General Gateway for New Users
Sandstone HPC: A Domain General Gateway for New Users
 

HybridAzureCloud

  • 1. Chris Condo ccondo@outlook.com August 20, 2014 Azure & HPC Configuration
  • 2. HPC = High Performance Computing • HPC is Microsoft’s “Big Compute” platform • Parallel and batch compute capabilities. • Applications: • Actuarial sciences. • Financial Risk Modeling • Digital content creation, transcoding. • Genetics research.
  • 3. • Head Node hosts a control panel which allows you to create cluster configurations. • The head node sends work the compute node clusters. • The work is split up into work units and executed in parallel. Typical on premise HPC configuration HEAD NODE ON PREMISE COMPUTE NODES
  • 4. Hybrid HPC-Azure configuration HEAD NODE ON PREM COMPUTE NODES AZURE COMPUTE NODES • Add capacity with worker nodes in the cloud. • Useful for scenarios such as seasonal bursts where extra capacity is needed occasionally. • Its not necessarily about speed, its about throughput.
  • 5. Azure Compute Node Size • Use either A8 or A9 class machines. • Problem sets can be configured to be processed in parallel across 8 or 16 cores. • These machines come at an added cost. • Managing that resource is important to controlling $$$ spent.
  • 6. Cluster Manager running on the head node
  • 7. Cluster Manager starting Azure Nodes. • There’s manual steps needed to start/stop jobs. • These steps can be automated via PowerShell
  • 8. Powershell can replace manual interaction. • Your application can be written to execute Powershell to add Azure capacity as needed. • Add nodes to your HPC Cluster dynamically • Get-HpcNodeTemplate $HpcNodeTemplate -Scheduler $Scheduler | Add- HpcNodeSet -Quantity $Quantity -Size $Size -Scheduler $Scheduler • Start or Stop the nodes • Get-HpcNodeTemplate $HpcNodeTemplate -Scheduler $Scheduler | Start- HpcNodeSet -Scheduler $Scheduler • Shut down the nodes when the job is done, don’t waste $$$ with a VM that’s not doing work.
  • 9. Windows Azure Management Libraries • A Natural .NET Framework that maps cleanly to the underlying REST API • Introduced late 2013 • Supports the Portable Class Library (PCL). • Ships as a NuGet Package you install from VS • Designed to use C# 5 async tasks • Easy to code with C# using Microsoft.WindowsAzure; using Microsoft.WindowsAzure.Management.Compute; using Microsoft.WindowsAzure.Management.Storage;
  • 10. Do cool stuff like create Azure VMs from a desktop. • Deploy cloud services from your .Net applications • Eg: Create a VM internal async Task CreateVM() { await _computeManagementClient.VirtualMachines.CreateAsync( “MyAzureService", “MyAzureDeployment", new VirtualMachineCreateParameters { RoleName = “MyName", RoleSize = VirtualMachineRoleSize.ExtraSmall }, new System.Threading.CancellationToken()); }
  • 11. Summary • HPC is specialized for certain industries that require massive computing power and can make use of parallel processing. • Azure “Burst to cloud” is an effective means to add capacity. • Automating the starting and stopping of your Azure resources can help control costs. • Powershell and/or WAML are 2 ways to develop operational scripts to automate Hybrid-HPC work management.