1. Microsoft Build 2017
11/1/2017 1:25 AM
Azure Batch
Media transcoding & pre-/post- processing
Rendering
Test execution
Monte Carlo simulations
Genomics
Deep Learning
OCR
Data ingestion, processing, ETL
R at scale
Compiled MATLAB
Engineering simulations
Image analysis & processing
Enable applications and algorithms to easily and efficiently run in parallel at scale
© Microsoft Corporation. All rights reserved. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

2. Batch workload characteristics
11/1/2017 1:25 AM
Batch workload characteristics
Job & Task Queue
Applications
Batch Pool
Persistent
Storage
INPUT
OUTPUT
Batch Task
© 2013 Microsoft Corporation. All rights reserved. Microsoft, Windows, and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries. The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

3. Low-Priority VMs Overview
Microsoft Build 2017
11/1/2017 1:25 AM
Low-Priority VMs Overview
Significantly lower priced compute:
Up to 80% discount compared to on-demand price - fixed price
All Batch VM sizes and regions
Uses surplus capacity; availability could vary; VMs could be preempted
Suitable workloads:
Batch processing, containerized applications, map/reduce
Distributed parallel jobs - many discrete tasks, interrupt tolerant, shorter task execution times, flexible job completion time
E.g. Dev, test – regression, scale, load
Value:
Get work done for lower cost, faster, or do more for same price
© Microsoft Corporation. All rights reserved. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

4. Low-Priority VMs Overview
Microsoft Build 2017
11/1/2017 1:25 AM
Low-Priority VMs Overview
Batch Pools:
Contain low-priority and/or dedicated VMs
Job does not need to know about pool VM types
Scale both VM types independently using resize operation or auto-scale
If preemption, pool automatically seeks to target to replace preempted VMs
// Create a Pool with mixture of dedicated and low-priority
CloudPool pool = PoolOperations.CreatePool (“Pool1”, “Standard_D14_v2”, vmConfig);
pool.TargetDedicated = 20;
pool.TargetLowPriority = 80;
pool.Commit();
// Get Pool info
CurrentDedicated = pool.CurrentDedicated;
CurrentLowPriority = pool.CurrentLowPriority;
// Resize Pool – no dedicated, all low-pri
pool.Resize(0, 100);
© Microsoft Corporation. All rights reserved. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

5. Options for using low-priority VMs
Microsoft Build 2017
11/1/2017 1:25 AM
Options for using low-priority VMs
Goal: Lowest cost
Pool config: All low-priority, no dedicated
Job: Most potential variation; available capacity may vary
Goal: Lower cost, guaranteed baseline capacity
Pool config: Fixed low-priority and fixed dedicated
Job: Min throughput and max job time guaranteed
© Microsoft Corporation. All rights reserved. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

6. Options for using low-priority VMs
Microsoft Build 2017
11/1/2017 1:25 AM
Options for using low-priority VMs
Goal: Minimal potential impact to current job time
Pool: Existing dedicated, supplement with low-priority
Job: Mainly faster and cheaper that all dedicated
Goal: Lowest cost, while maintaining capacity
Pool: All low-priority, set dedicated = preempted
Job: Reduced impact to job time, capacity fairly consistent
© Microsoft Corporation. All rights reserved. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

7. Auto-scaling in action
Microsoft Build 2017
11/1/2017 1:25 AM
Auto-scaling in action
Pool auto-scale formula – maintain 500 nodes:
numPreemptedCount =
avg($PreemptedNodeCount.GetSample(6 * TimeInterval_Minute));
$TargetLowPriorityNodes = 500;
$TargetDedicatedNodes = numPreemptedCount;
© Microsoft Corporation. All rights reserved. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.