We present applications of Azure Services such as Azure IaaS/PaaS and Azure RemoteApp in computational fluid dynamics and sparse linear algebra. We also present Microsoft Machine Learning Studio in prediction of the heating load in the buildings.

1. Microsoft Innovation Center for Technical
Computing
MICROSOFT AZURE IN HPC SCENARIOS
Lukasz Miroslaw, Ph.D.
lukasz.miroslaw@hsr.ch
18.11.2015, MICROSOFT SWITZERLAND

2. Challenges
2
57 % of users are dissatisfied with their desktop
computing capacity*
* Source: US Council of Competitiveness: http://www.compete.org, theubercloud.com
Computing: too slow
Memory: too small
Fig. Sometimes solving a problem with IT is hard.

3. Challenges
3
$70,000 server => $1M cost over 3 years
High costs of IT infrastructure

4. Low Cost of running in the cloud
 Cost Model assumes that the hardware makes 7% of Total Costs
4
Fig. Cost of a GFLOP in U.S. Dollars on different Microsoft Azure
nodes and a private HSR cluster.

5. Motivation
5

6. Agenda
6
 Use Case #1: Remote physical simulations with external partners.
 Use Case #2: Scale out physical simulations in the cloud.
 Use Case #3: Stellar Classification, Prediction of Energy Efficiency in
buildings
 Conclusions.

7. Agenda
7
 Use Case #1: Remote physical simulations with external partners.
Azure IaaS, Remote App, Azure Batch
 Use Case #2: Scale out physical simulations in the cloud.
SimplyHPC, HPC Pack
 Use Case #3: Stellar Classification, Prediction of Energy Efficiency in
buildings
AzureML
 Conclusions.

8. What is Computational Fluid Dynamics?
 CFD is the science to simulate fluid flow, heat and mass transfer and
chemical reactions
8

9. What is Computational Fluid Dynamics?
 Airflow simulation around sky-diving Santa Claus.
9
* Source: Desktop Engineering

10. Use Case #1: Collaborative Simulations of Electrical Arcs
10

11. Use Case #1: Collaborative Simulations of Electrical Arcs
 Goal #1: Develop a Cloud-based algorithm for electrical arc simulation
 Microsoft Azure Research Award in 2014
 Contact: Kenji Takeda (Microsoft Research)
 Goal #2: Provide simulation tool to partners in Brasil and Deutschland
 Ongoing collaborations
 Streamer International (CTI Project)
 Panasonic
 Fraunhofer SCAI
 WEG
11

12. 1st Use case: Instant ANSYS
12
 VM:
 D14 with 16 core
CPU, 112 GB RAM,
Windows Server
2012
 MpCCI, ANSYS
preinstalled
 Storage: locally
redundant,
automatically
scallable
 License Server (LS)
on A0 in Germany
Customer
VM
LS

13. INSTANT ANSYS
13
No Installation. No configuration. No up-front costs.
 Access to powerful VMs with ANSYS already preinstalled
and preconfigured.
 Access to redundant and highly available storage.
 Disaster Recovery and 99.5% SLA.
 Connection to on-premise infrastructure with IPSec VPN.

14. Use Case #1: INSTANT ANSYS
14
IaaS DEMO

15. 2nd Use case: Linux VM
15
 The UberCloud: Making Technical
Computing available in the Cloud
 UberCloud Community:
 +2500 companies and
individuals:
 +60 cloud providers,
 +80 software providers,
 several hundred consulting
firms and individual experts.
 OpenFOAM added to Azure
Marketplace
 Docker containerization
www.ubercloud.com

16. 2nd Use case: Linux VM
16
DEMO
The compute environment you ordered is now
ready.
Access your compute environment via remote
desktop connection (Chrome 8+, Firefox 7+,
Opera 11+, IE 9+)
Launch
Your password for remote desktop access is:
TN1b39pv4Djw

17. Azure RemoteApp
17
Deliver apps from the cloud, cost-
effectively
Simplify your infrastructure
Run Windows apps anywhere
Centralize your apps, help secure your data

18. Azure RemoteApp
18
Windows applications as a service
accessible from anywhere.

19. Costs
19
 VM with 16 cores and 56 GB RAM costs 2.11 CHF / hour (D14)
 1 TB of Storage costs 30 CHF / month
 RemoteApp starting price: $10 / user / month (40h included)
 Online Calculator
Azure in Education
Faculty will receive a 12 month,
$250/month account
Students will receive a 6 month,
$100/month account

20. Short Summary
20
 + Powerful VMs that can be started/stopped on-demand increase
the productivity in our group.
 + Virtual images with OS and different software version to avoid
problems with backward compatibility.
 + Students and team members can manage their own VMs and
reduce the costs of support.
 - Storage File Service can be easily mapped to a drive on the VM
but not on premises.
 - Only a single user can access one VM.

21. Scalability Tests on Microsoft Azure

22. HPC Pack IaaS Demo

23. SimplyHPC: Light-weight Cloud Orchestrator for MS Azure
 What is SimplyHPC?
Framework
23
SimplyHPC:
1) Distributed framework for
Microsoft Azure,
2) Set of PowerShell scripts.

24. SimplyHPC = Simpler Deployment

25. Performance and Scalability
 Example #1: Solving linear systems with PETSc and HPCG
25
Fig. Performance in GFlops of PETSc solving ruep (right) matrix
system and HPCG Benchmark (left) on different Microsoft Azure
nodes and a private HSR cluster.

26. Performance and Scalability
26
 Example #2: ANSYS CFX

27. Performance and Scalability
27
Fig. Strong scaling of ANSYS CFX of the compressor (11 mln nodes).
 Example #2: ANSYS CFX

28. Azure Batch
28
Batch is a managed service for batch
processing or batch computing - running a
large volume of similar tasks to get some
desired result.

29. Short Summary
 SimplyHPC: a framework to simplify cluster deployment and
job submission.
 Set of light-weight PowerShell scripts to submit, execute and
monitor multi-threaded jobs on Windows Azure.
 Easy to use. No cloud-related knowledge necessary.
 Run the jobs from command line and download the results
directly to your Azure Storage.
 Up to 9x faster than native MS HPC Pack scripts.
 Available at https://github.com/vbaros/SimplyHPC
29
L Miroslaw, V Baros, M Pantic, H Nordborg, Unified Cloud Orchestration
Framework for Elastic High Performance Computing on Microsoft Azure,
NAFEMS World Congress 2015

30. Short Summary
30
 + Scaling properties of Microsoft Azure is comparable to the on-premises
cluster.
 HSR Cluster: 7.3 days (176 hours), limited availability.
 Microsoft Azure: 4.9 days (118 hours), ca. 50% faster, 100% availability.
 + Dynamic scaling (up- / downscaling) and instant access to the newest
hardware reduces the costs.
 + (Un)limited computing at competitive price.
Cluster composed of 32 x A8 nodes (=256 cores) costs
32 x 2.11 CHF/h = ca. 68 CHF/h
 - Upscaling > 100 cores should be planned in advance.

31. Microsoft Azure Machine Learning Studio
 Three types of knowlege:
 Know-What (facts)
 Know-How (processes)
 Know-Why (reasons)
31
Image credit: Univ. Hamburg

32. AzureML Studio
 Key goals of Machine Learning:
 Prediction
 Classification
 Clustering
 Collaborative Filtering
32
Image credits: OpenCV, Snipview, Stanford

33. AzureML: Stellar Classification
 Classification Challenge:
 HYG database* is a compilation of
of stellar data from three main
catalogues.
 Contains ca. 120k stars, 37
spectral characteristics.
 2D classification scheme based on
temperature (color index) and
brightness (absolute magnitude).
 Data is incomplete and may
contain a few misclassifictions.
 Prediction Engine developed in
AzureML
33
Credits: Michael Pantic (HSR)* http://www.astronexus.com/hyg

34. AzureML Example: Heating Load Prognosis
34
Image credit: SAB Magazine
Input:
- Roof area
- Overall hight
- Glazing area
- Surface area
- ...
Output:
- Heating load
prediction

35. AzureML Workflow
35
Machine Learning Workflow
1. Hypothesis
2. Data Preparation
3. Model
4. Test
5. Evaluate
A. Tsanas, A. Xifara: 'Accurate quantitative estimation of energy performance of
residential buildings using statistical machine learning tools', Energy and Buildings,
Vol. 49, pp. 560-567, 201
- 8 physical characteristics
from 768 buildings
- Goal: predict buldings’
heating load and cooling
load
- Architects need to compare
several building designs
before selecting the final
approach

36. AzureML: Cost Model
36

37. AzureML: Short Summary
37
 Very fast prototyping. Load the system with data, test different
Machine Learning methods.
 Platform for Internet of Things: Event Hubs, Stream Analytics.
 Share the models & results.
 Deploy web services fast.
 Develop own methods in Python and R Statistics.

38. Summary
38
 Computing and storage at competitive
price.
 High Availability, data redundancy,
disaster recovery services are included.
 Data transfer take some time.
 Up- and downscaling resources
dynamically. Higher productivity.
 „Cloudify” your system’s complexity.