umeå university cloud computing...
TRANSCRIPT
Umeå University Cloud Computing Research Jakub Krzywda Umeå University [email protected] Poznań 2014-03-18 www.cloudresearch.org
Erik
Elm
roth
, el
mro
th@
cs.u
mu.
se
Studia
2
Jaku
b Krz
ywda
, ja
kub@
cs.u
mu.
se
• Praca inżynierska – Platforma udostępniania danych
krytycznych w architekturze SOA – dr Jacek Kobusiński
• Indywidualny Tok Studiów – Rozproszone przetwarzanie danych – prof. Jerzy Brzeziński
Erik
Elm
roth
, el
mro
th@
cs.u
mu.
se
Praktyki w Inria
3
Jaku
b Krz
ywda
, ja
kub@
cs.u
mu.
se
• kwiecień – wrzesień 2013 • Inria Grenoble • grupa STEEP – zrównoważony
rozwój • projekt – miary eksurbanizacji
Raport: http://hal.inria.fr/hal-00907081
Erik
Elm
roth
, el
mro
th@
cs.u
mu.
se
Doktorat na uniwersytecie w Umeå
4
Jaku
b Krz
ywda
, ja
kub@
cs.u
mu.
se
• od listopada 2013
• Umeå University
Department of Computing Science
Cloud and Grid Computing Group
Erik
Elm
roth
, el
mro
th@
cs.u
mu.
se
Table of Contents
5
Jaku
b Krz
ywda
, ja
kub@
cs.u
mu.
se
• Introduction to Cloud Computing
• Cloud & Grid Computing Group
@ Umeå University
• Projects:
– Cactos
– Cloudberry
Erik
Elm
roth
, el
mro
th@
cs.u
mu.
se
What is Cloud Computing?
New term for a long-held dream of computing as a utility. Armbrust, Michael, et al. "Above the Clouds: A Berkeley View of Cloud Computing." (2009).
Model for enabling access to a shared pool of computing resources. Mell, P., & Grance, T. (2009). The NIST definition of cloud computing. National Institute of Standards and Technology, 53(6), 50.
7
Jaku
b Krz
ywda
, ja
kub@
cs.u
mu.
se
Erik
Elm
roth
, el
mro
th@
cs.u
mu.
se
Essential Characteristics
• On-demand self-service
• Broad network access
• Resource pooling
• Rapid elasticity
• Measured service
8
Jaku
b Krz
ywda
, ja
kub@
cs.u
mu.
se
Mell, P., & Grance, T. (2009). The NIST definition of cloud computing. National Institute of Standards and Technology, 53(6), 50.
Erik
Elm
roth
, el
mro
th@
cs.u
mu.
se
Deployment Models
• Public cloud
• Private cloud
• Community cloud
• Hybrid cloud
9
Jaku
b Krz
ywda
, ja
kub@
cs.u
mu.
se
Erik
Elm
roth
, el
mro
th@
cs.u
mu.
se
Service Models
• Infrastructure as a Service (IaaS) • Platform as a Service (PaaS) • Software as a Service (SaaS)
10
Jaku
b Krz
ywda
, ja
kub@
cs.u
mu.
se
http://www.cloudsherpas.com/cloud-services/google/google-cloud-platform/ http://developers.google.com/cloud/
Erik
Elm
roth
, el
mro
th@
cs.u
mu.
se
Senior researchers Post docs
Erik Elmroth, Professor
Francisco Hernandez Assistant Professor
Johan Tordsson, Assistant Professor
P-O Östberg, Researcher
Cristian Klein, PhD
Luis Tomás, PhD
Daniel Espling, PhD
PhD students
Ahmed Ali-Eldin, PhLic
Jakub Krzywda
Ewnetu Bayuh Lakew, PhLic
Wubin Li, PhLic
Mina Sedaghat, PhLic
Petter Svärd, PhLic
Kosten Selome Tesfatsion
Gonzalo Rodrigo
PhD students Others
Amardeep Mehta
Olumuyiwa Ibidunmoye
Peter Gardfjäll, Sys. Dev, PhLic
Lars Larsson, System developer
Lennart Edblom, Senior lecturer
Tomas Forsman, Systems expert
Niclas Lockner, Research assistant
www.cloudresearch.org 13
Jaku
b Krz
ywda
, ja
kub@
cs.u
mu.
se
Erik
Elm
roth
, el
mro
th@
cs.u
mu.
se
Scope of Umeå University Cloud Research
Our focus – Autonomic capacity
management – Throughout service lifecycle – Single and multiple
datacenters
14
Jaku
b Krz
ywda
, ja
kub@
cs.u
mu.
se
15
Static, energy consuming, inflexible, VM-centric hosting in data centres
Dynamic, energy efficient, elastic, Service-centric hosting in the cloud
Context-‐Aware Cloud Topology Op5misa5on and Simula5on h;p://cactosfp7.eu
A very short view on CACTOS Partners Umeå Universitet, SE Ulm Universität, DE REALTECH AG, DE The Queen’s University of Belfast, UK Flexiant Limited, UKFZI Forschungszentrum Informa5k, DE Dublin City University, IR Dura+on: Oct 2013 – September 2016 Total cost: 4,761,232 €
Why CACTOS?
19
• Data Centre are built with x86 single core CPUs
• Applica5on too slow? à Buy new HW
Good old days
• Mul5-‐Core CPUs to address energy challenge
• X86 offers begin to differ and specialised processors emerge
• Many network op5ons
• App too slow? à Change your SW
The recent past up to now
• Heterogeneous CPU/APU
• Lots of network op5ons
• App to slow? à Choose the right architecture!
Near Future
Cactos in a nutshell
Data Centre Operators/Cloud Operators
CactoScale
analyze datalogs
collectapplication
behavior data
collect infrastructure and
hardware data
Cloud Middleware Developers, Cloud Infrastructure Providers,
Data Centre Operators
CactoOpt automatic mapping of workloads
determine best fitting resource
find most appropriate provider
Cloud Middleware Developers, Cloud Infrastructure Providers,
Data Centre Operators
CactoSim
simulate optimization
models
predict behavior of applications on
different resources
validate and improve models
20
CactoOpt Architecture
21
Cluster level • resource utilization • load mixing Data center level • proactive plan • multiple criteria / objective functions
Erik
Elm
roth
, el
mro
th@
cs.u
mu.
se
Why Cloudberry?
23
Jaku
b Krz
ywda
, ja
kub@
cs.u
mu.
se
• Vendor Lock-In – Data – Software
• Scalability – Support billions or trillions of devices
• Workload distribution – Over clusters, data centers, countries
Erik
Elm
roth
, el
mro
th@
cs.u
mu.
se
Cloudberry Architecture 1. Initial filtering
phase 2. Refined filtering
phase 3. Ranking phase 4. Deployment
phase
26
Erik
Elm
roth
, el
mro
th@
cs.u
mu.
se
Cloudberry Architecture
27
Jaku
b Krz
ywda
, ja
kub@
cs.u
mu.
se
Erik
Elm
roth
, el
mro
th@
cs.u
mu.
se
How to choose the best resource?
28
Jaku
b Krz
ywda
, ja
kub@
cs.u
mu.
se
• Overhead of virtualization • Benchmark resource requirments • Forecast workload • Transformation application
requirements and resource usedge