Welcome! Erik Elmroth Facts and figures Swedens fifth oldest - - PDF document

1

Welcome!

Erik Elmroth

Facts and figures

• Sweden’s fifth oldest

University

• Established 1965
• 31 500 students
• 4 300 employees
• Umeå has 120 000 inhabitants

2

UmU Campus close-up UmU Campus

3

Why does WASP include Cloud Computing?

• As a research topic:
• The clouds should be autonomous themselves
• As an education topic:
• (See above)
• Important foundation for a lot of autonomous

systems and software development

Why is this course module given at UmU?

• UmU has been among the European leaders in

cloud computing (and cloud management) research since 2006

• Sample highlights:
• RESERVOIR 2008-2011 – EU’s first and largest cloud

computing project

4

Why is this course module given at UmU?

• UmU has been among the European leaders in

cloud computing (and cloud management) research since 2006

• Sample highlights:
• RESERVOIR 2008-2011 – EU’s first and largest cloud

computing project

• 7 EU FP7 & H2020 projects on cloud topics
• Cloud Control – 2nd largest VR framework grant ever
• Cloud Control Workshops
• Northern Sweden becoming a datacenter

metropole?

Facebook datacenter in Luleå

5

Autonomous Cloud Research at UmU

• Large team, focused scope
• Highly multi-disciplinary

collaborations

• Target infrastructures…
• individual servers
• clusters
• rack-scale systems
• full datacenters
• distributed edge clouds

6

Deep Industrial Collaborations

www.cloudresearch.org

Distributed Systems (Autonomous cloud management)

• Ahmed Ali-Eldin, UMASS, Amherst, USA
• Gonzalo Rodrigo, Lawrence Berkeley Lab, USA
• Mina Sedaghat, Accenture, Stockholm
• Luis Tomás, Red Hat, Spain
• Wubin Li, Ericsson Research, Montreal
• Lei Xu, IBM, Dublin
• Daniel Bergström, Ericsson Research, Luleå

7

Recent PhD theses

• 1. G. Rodrigo. HPC Scheduling in a Brave New World (2017)
• 2. M. Sedaghat. Cluster Scheduling and Management for Large-Scale

• 3. A. Ali-Eldin. Workload characterization, controller design and

• 4. E.B. Lakew. Autonomous cloud resource provisioning: accounting,

• 5. W. Li. Algorithms and Systems for Virtual Machine Scheduling in Cloud

• 6. P. Svärd. Dynamic Cloud Resource Management: Scheduling, Migration

• 7. D. Espling. Enabling Technologies for Management of Distributed

• 8. P-O. Östberg. Virtual infrastructures for computational science:

• 9. J. Tordsson. Portable Tools for Interoperable Grids: Modular

Cloud Control Workshops

• Foster multi-disciplinary

research on Cloud Management, leveraging

– autonomic computing, control theory, distributed systems, energy management, machine learning, mathematical statistics, performance modeling, systems management, etc.

• Flexible format

– Often mainly discussion-oriented

• (Surprisingly) easy to attract participants

– Key people from, e.g., Google, VMware, Intel, Ericsson, Netflix, IBM and academia world-wide

8

9th Cloud Control Workshop (last summer)

• 61 invited participants

– 2 Keynotes – 16 Short presentations (15+5 mins) – 7 Even shorter young presentations (10+5 mins) – 16 discussion sessions, 1 hour each cloudresearch.org/workshops/9th

Course team

• Dr. Cristian Klein (main organizer)
• Assoc. Prof. Johan Tordsson (lecture)
• Prof. Johan Eker (lecture + testbed)
• Dr. Ewnetu Bayuh Lakew (tutorial/assignments)
• PhLic. Olumuyiwa Ibidunmoye “Muyi” (tutorial/

assignments)

• Prof. Erik Elmroth (overall responsibilities)

9

Keynote Speakers (Today 10.30-12)

Filip Grządkowski, Google Engineering manager “Automated container deployment, scaling, and management with Kubernetes”

Keynote Speakers (Tomorrow 8-9.30)

Joel Segerlind, Burt Product owner “Moving from pets to cattle at Burt”

10

Additional course highlights

Assignments

• Aligned with lectures
• In groups mixing software students with control students
• Skeleton implementation of a real-world application
• In essence extending the tutorial
• Scalable and resilient video conversion service
• Hands-on exposure to both software and control challenges

(motivating the student mix)

• On the latest and greatest Ericsson Cloud

HPC2N Visit: Not a cloud data-center, but gives a good impression about the scale, energy usage and cooling needs of a modern cloud datacenter

Ericsson RESEARCH Data Center

WASP

11

• OpenStack based
• Launch virtual machines

(Linux)

• Create virtual networks
• Deploy services
• Collect & store data

#1 Infrastructure-as-a-service

#2 Analytics-as-a-service

• Work with WASP/WARA data

collected and stored at ER DC

• Provide environment for

analysis/machine learning/etc.

• Hadoop MapReduce, Hadoop

Spark, Apache Zeppelin

#3 Edge Cloud Platform

• Move computations around to

decrease latency, improve performance, etc.

• Monitor cloud behaviour

Practicalities

• Lecture room
• This room, all lectures
• Computer labs
• MIT building, top floor (MA416, MA426, MA436, MA446)
• Lunches
• Restaurant Hjortron at Universum (2nd floor)
• Coffees
• Outside lecture room or near computer lab & HPC2N
• WiFi
• Eduroam should do it. Get in touch if not.

12

Dinner at Rex, today at 18.30 Dinner at Rex, today at 18.30

13

Dinner at Rex, today at 18.30

Juliette: 1920th cocktail bar (entrance to our reserved room)

All in all…

Our ambition is that you should

• understand what clouds are and can do for you
• gain experience in building cloud applications
• learn how clouds should be built and why they

should be autonomous … and hopefully get excited about it ... and have a great time among colleagues ...

14

Enjoy!