In this deck from ISC 2015, Wolfgang Gentzsch from The UberCloud presents: HPC Workloads Using Docker.
"The UberCloud Project provides an environment to run HPC workloads on a variety of cloud providers. Wolfgang elaborates on the projects and provides his experience about the roadblocks towards a more flexible cloud-backed HPC future."
Watch the video presentation: http://insidehpc.com/2015/07/video-hpc-workloads-using-docker/
Sign up for our insideHPC Newsletter: http://insideHPC.com/newsletter
2. 2012: UberCloud started
its free voluntary Cloud Experiments
HPC as a Service, on demand, in a team experiment
to explore the end-to-end process
using cloud computing resources
as a service, on demand, at your fingertips
and learning how to resolve the roadblocks.
TODAY:
172 experiments, 2500 companies, 72 countries, 50 case studies
Supported by our sponsor INTEL
3. Major Cloud Roadblocks
Supply
Cloud providers
ISVs
Consultants
Trainers
Demand
Engineers
Scientists
Data analysts
Experts
.
.
.
.
.
Complexity
Data
Transfer
SecurityLicensing
Uncertain
Cost
Roadblocks
4. 2015: The UberCloud Marketplace
Supply
Cloud providers
ISVs
Consultants
Trainers
…
Demand
Engineers
Scientists
Data analysts
Experts
Marketplace
And HPC Docker
7. UberCloud application containers
Ready-to-execute packages of software, designed to deliver the tools that
an engineer needs
Ready to execute, in an instant. No need to install software, deal with
complex OS commands, or configure.
The ISV or Open Source tools are pre-installed, configured, and tested, and
are running on bare metal, without loss of performance.
Based on open Linux container technology
Enhanced for engineering & scientific applications
ANSYS, CD-adapco, OpenFOAM, Gromacs, Scilab, and more
8.
9. Virtual Machines vs Docker
VM: Each virtualized application includes not only the application and the necessary binaries and libraries, but also
an entire guest operating system.
Docker Engine container comprises application and its dependencies. Isolated process in user space on host
operating system, sharing kernel with other containers. Enjoys resource isolation and allocation benefits ofVMs but
more portable and efficient.
9
10. Builder
Launcher
Controller
ISV DataTools
Stackable units with tools (ex: encryption), ISV application codes (ex: OpenFOAM).
Just add your own codes and data.
Run anywhere with UberCloud Run Time.
Scale up or down the compute power as needed.
Collect granular usage data, logs.
Monitor, alert, report.
Any Workstation Any Cluster Any Cloud
Run Time Run Time Run Time
Containers: Build once, run anywhere
11. Builder
Launcher
Controller
ISV DataTools
Stackable units with tools (ex: encryption), ISV application codes (ex: OpenFOAM).
Just add your own codes and data.
Run anywhere with UberCloud Run Time.
Scale up or down the compute power as needed.
Collect granular usage data, logs.
Monitor, alert, report.
Any Workstation Any Cluster Any Cloud
Run Time Run Time Run Time
Containers: Build once, run anywhere
12. Builder
Launcher
Controller
ISV DataTools
Stackable units with tools (ex: encryption), ISV application codes (ex: OpenFOAM).
Just add your own codes and data.
Run anywhere with UberCloud Run Time.
Scale up or down the compute power as needed.
Collect granular usage data, logs.
Monitor, alert, report.
Any Workstation Any Cluster Any Cloud
Run Time Run Time Run Time
Containers: Build once, run anywhere
13. Builder
Launcher
Controller
ISV DataTools
Stackable units with tools (ex: encryption), ISV application codes (ex: OpenFOAM).
Just add your own codes and data.
Run anywhere with UberCloud Run Time.
Scale up or down the compute power as needed.
Collect granular usage data, logs.
Monitor, alert, report.
Any Workstation Any Cluster Any Cloud
Run Time Run Time Run Time
Containers: Build once, run anywhere
14. Containers remove Portability & Standard related challenges.
By integrating 3rd party tools we tackle the following:
16. OpenFOAMTest: Bare Metal vs Docker
PerformanceTest Results demonstrating comparison between the same OpenFOAM
run (with 1.4 million cells) repeated on bare metal and in a Docker container.
18. UberCloud multi-container environment
Intel MPI libraries pre-setup for multi-node high performance
Intel compilers achieve architecture specific high performance
Networking between containers
Setting up shared file system access (NFS)
Secure communication (ssh) between containers
Benefits from single-node transferred to multi-node containers:
No setup of the environment
Portability and packaging
Easy access and use
Low overhead
+
19. Hot Spot Question for Jérôme Petazzoni
Docker recently announced that it acquired SocketPlane to help with container
networks. I am very interested in some more background and details and how
this could support HPC multi-container approach.
There are multiple container implementations, Docker, CoreOS and now a
collective effort, runC. Can you talk about what this all means for the community.
Virtualization has been around for a long time, why re-invent the wheel?
In HPC clusters shared file systems are quite commonly used. Can you comment
on how to secure a Docker installation in conjunction with a shared file system.
Wouldn't the shared file system be vulnerable when Docker is in the mix?
20. ANSYS Container Demo
ANSYS multi-physics (5 s/w) running inside Docker container
Pre- and post-processing in the cloud works the same way that it would work
on the engineer’s desktop
User experience in the cloud matches that on the desktop
GPU support for high-resolution remote visualization
Automated monitoring keeps you connected, under control
4-min demo: https://goo.gl/YIrTr4
6-min demo: http://goo.gl/pDXiTq
21. Containers as basic technology
for the UberCloud SaaS Factory
UberCloud SaaS
Factory
Software SaaS
Commercial ISV
Open Source
In-house
Public Cloud
Private Cloud
On-Premise
$$$
We turn high-TCO, complex software into user-friendly pay-per-use services:
Development Services (Docker)
--------------------
Deployment & Management
Services
------------------
Marketplace
Providers Consumers
$
UberCloud SaaS Factory