Status, trends and challenges in cloud and multi-cloud service provisioning Erik Elmroth Department of Computing Science & HPC2N Umeå University www.cloudresearch.se
From where I view the clouds Three main projects (FP7 IPs) Introduced federated clouds. Among EUs first major cloud project. Optimized cloud services over the complete lifecycle. Non-functional aspects. Pioneering federated storage clouds. Raised abstraction level. Media- & telecom apps.
Critical performance requirements - to be cost-efficiently met Extremely rapid growth (from global scale) YouTube (16 months) 100 mil/movies per day, 20 mil. unique users per month AppStore (19 months): Over 100000 Iphone programs, over 3 billion downloads Regular/planned peaks Banks, tax filing Market campaign effects Unexpected peaks News related video streaming Stock trading peaks at financial crises Regional aspects in usage patters Regional concerns (news, events, etc) Time-dependent usage-patterns
Critical performance requirements - to be cost-efficiently met Extremely rapid growth (from global scale) YouTube (16 months) 100 mil/movies per day, 20 mil. unique users per month AppStore (19 months): Over 100000 Iphone programs, over 3 billion downloads Regular/planned peaks Banks, tax filing Market campaign effects Unexpected peaks New related video streaming Stock trading peaks at financial crises Regional aspects in usage patters Regional concerns (news, events, etc) Time-dependent usage-patterns
Erik Elmroth elmroth@cs.umu.se New data center challenges Traditionally: managed peak loads by hosting too much hardware Requirements for an elastic data center infrastructure Google @ The Dulles, OR Today s Clouds provide partial solutions
NIST definition of cloud computing 5 characteristics On-demand self-service Broad network access Resource pooling Rapid elasticity Measured service National Institute of Standards and Technology 3 service models Software-as-a-Service Plattform-as-a-Service Infrastructure-as-a-Service 4 deployment models Public Private Community Hybrid
Next Generation Clouds What? Large-scale IT capacity Compute + storage + network Automatically increase, decrease & migrate Large-scale management How? Virtual machines -Abstracts hardware Grid technology -Distributed virtual resource Business Service Management - Dynamic SLA management Autonomic systems
Erik Elmroth elmroth@cs.umu.se Service Provider Bursted Private Clouds THREE BASIC SCENARIOS Private infrastructure Infrastructure Provider Service Provider Federated Clouds Infrastructure Provider Infrastructure Provider Infrastructure Provider Service Provider Broker Multi-clouds Infrastructure Provider Infrastructure Provider Infrastructure Provider
Some trends Service types Infrastructure platform & software Infrastructure types Compute or storage unified capacity management Abstraction level Focus on data, not storage Focus on work done, not computers Non-functional aspects Service qualities. Not what can be done but how good it can be done Location For performance, legal, trust or economical reasons Systems scale Autonomous systems Information scale Data & metadata. User, services and systems data Security & Trust The most common areas for questions and concerns 11
Erik Elmroth elmroth@cs.umu.se
Erik Elmroth elmroth@cs.umu.se Senior researchers Project admin. Erik Elmroth, Professor Francisco Hernandez, Assistant Professor Johan Tordsson, Assistant Professor Lei Xu, Post Doc P-O Östberg Post Doc Lennart Edblom, Senior lecturer PhD students Ahmed Ali- Eldin Daniel Henriksson Ewnetu Bayuh Lakew Wubin Li Mina Sedaghat Petter Svärd Systems developers/systems experts cloudresearch.se Tomas Forsman, Systems expert Peter Gardfjäll, Systems Developer Sebastian Gröhn, Research assistant Anders Häggström, Research assistant Lars Larsson, Systems Developer
14
So far, lots of focus on compute clouds Data is everywhere and growing Data is key to cloud management Unified management view for compute, storage and network Larger systems More data Locality awareness More focus on self-managed systems Split-deployment dependencies interoperability Managing information Information for management Managing capacity (resources) Managing services Virtualizing resources & abstracting storage 15
Erik Elmroth elmroth@cs.umu.se Nåt om from. amazon-type to largescale distributed. data + storage+ netowrok management for.
Erik Elmroth elmroth@cs.umu.se Nån typ av allmän intro till vad cloud är Kanske med lite om roller, virtualisering, data centers Management with respect to non-functional apsects/requirements Få med alla möjligt OPTIMIS-aspekter TREC Legal aspects
Managing capacity (resources) Managing services Virtualizing resources & abstracting storage 18
Erik Elmroth elmroth@cs.umu.se Cloud Resource Management For whom? Service providers Infrastructure providers What? Compute + storage + network Low and high level management How? Single abstraction multiple use (scenarios) General tools for key functionality Flexibility in deployment and configuration
Example (low level management): Elasticity- & access control Elasticity control Control the system s handling peaks & lows Inceasing ability to meet SLAs Reduces resource consumption Access control Overbooking of elastic services Access control quality directly determines income and SLA violation rate
Erik Elmroth elmroth@cs.umu.se Holistic cloud management Business Level Objectives Management constraints Monitor Monitor Algorithms Policies Algorithms Policies Algorithms Policies Algorithms Policies Algorithms Policies Algorithms Policies
Live migration Transfer a VM from one host to another without disrupting services. UMU:s research: better algorithms for live migration Caching, compression Page priority Improves: Resource utilization Reliability Flexibility Enables cloud management!