t-110.5121 mobile cloud computing basics 14.09 · what is cloud computing 1. the illusion of...
TRANSCRIPT
9/14/2011
Yrjö Raivio
Aalto University, School of Science
Department of Computer Science and Engineering
Data Communications Software
Email: yrjo.raivio(at)aalto.fi
Course email: t-110.5121(at)tkk.fi
© Y Raivio
T-110.5121 Mobile Cloud Computing
Basics
14.09.2011
© Y Raivio
• Definition
• Key benefits
• Technologies
• Challenges and opportunities
• Conclusions
Outline
9/14/20112
Lecture is based on the Virtual Keynote by Doug Terry: ”Technology in the Cloud – Plus
some Challenges and Opportunities”, Microsoft Research, June 1, 2011, available from
http://techpack.acm.org/cloud/
© Y Raivio
Definition
9/14/20113
“Mobile Cloud computing is a model for enabling convenient, on-demandmobile network access to a shared pool of configurable mobilecomputing resources (e.g., networks, servers, storage, applications, and services) that can be rapidly provisioned and released with minimal management effort or service provider interaction.”
Adapted from: P. Mell and T. Grance, “The NIST Definition of Cloud Computing”, 2009
© Y Raivio
9/14/20114
What is cloud computing
1. The illusion of infinite computing resources available on demand, thereby eliminating the need for Cloud Computing users to plan far ahead for provisioning.
2. The elimination of an up-front commitment by Cloud users, thereby allowing companies to start small and increase hardware resources only when there is an increase in their needs.
3. The ability to pay for use of computing resources on a short-term basis as needed (e.g., processors by the hour and storage by the day) and release them as needed, thereby rewarding conservation by letting machines and storage go when they are no longer useful.
Source: Ambrust et al, ”Above the Clouds: A Berkeley View of Cloud Computing”, 2009
© Y Raivio
• ”The horse is here to stay,
but the automobile is only a
novelty - a fad”, President of
the Michigan Savings Bank,
1903
• “..we expected to get orders
for five machines, we came
home with orders for 18.”,
Thomas Watson, Jr., April 28,
1953
• “There is no reason for any
individual to have a personal
computer in their home.”
Ken Olsen, President, Digital
Equipment Corp., 1980
History
9/14/20115
Sources: Joe Sherlock, The View Though The Windshield, available at: http://www.joesherlock.com/nwsltr1.html ;
Rolf Harms and Michael Yamartino: The Economics of the Cloud, Nov. 2010.
© Y Raivio
9/14/20116
• “The interesting thing about Cloud Computing is that we’ve redefined Cloud Computing to include everything that we already do. . . . I don’t understand what we would do differently in the light of Cloud Computing other than change the wording of some of our ads.”
Larry Ellison, Wall Street Journal, September 26, 2008
• “It’s stupidity. It’s worse than stupidity: it’s a marketing hype campaign. Somebody is saying this is inevitable — and whenever you hear somebody saying that, it’s very likely to be a set of businesses campaigning to make it true.”
Richard Stallman, The Guardian, September 29, 2008
Some skepticism
© Y Raivio9/14/2011
Key benefits
© Y Raivio
Economies of scale
9/14/20118
Source: Rolf Harms and Michael Yamartino: The Economics of the Cloud, Nov. 2010.
• Cheaper MIPS (5-7 times)
• Better utilization of computing resources (5-10% to 60-80%)
• Multi-tenancy: one instance can serve several customers
• Less admin people per server (from 1:100 up to 1:10 000)
• Worth 1$ IT requires 8$ admin costs
© Y Raivio
9/14/20119
Elasticity
Source: Ambrust et al, Above the Clouds: A Berkeley View of Cloud Computing, Feb 2009
© Y Raivio
9/14/201110
Resource planning
• Resources can be optimized to meet service needs
• Service integration time can be shortened, example SMSC setup from 2 weeks to 4 minutes
Time
Turnover
Private
cloud
Public
cloudHybrid
cloud
Early
development
Exponential
growth
Mature
market
Source: Rolf Harms and Michael Yamartino: The Economics
of the Cloud, Nov. 2010.
© Y Raivio
Pay-as-you-go
9/14/201111
• Avoid high upfront investment, avoid risk
• Adapt to changing business
• Buy or lease?
• Amortizise value to investment period
• Equated Monthly Installment
• Net Present Value
1)1(
)1(
12
1212
mr
mrr
m EA
Er
S
r
CPNPV
N
N
T
TT
)1()1(0
Private cloud
Public cloud
a
b
F (cost)
Source: TeliaSonera
b a
b
dtyfByfA0
)()(
• Minimize
• www.cloudonomics.com
© Y Raivio
9/14/201112
Realism – Case SMSC
Source: Y. Raivio, O. Mazhelis, K. Annapureddy and P. Tyrväinen, Hybrid Cloud Architecture for
Short Message ServicesSAC2012, submitted for approval
• Granularity
• Public cloud charging based on full hours
• Startup takes ~ a minute
© Y Raivio
Always available
9/14/201113
• Anyone, anytime, anywhere
• High availability?
• Amazon Elastic Compute Cloud (EC2)
• 99.95% = max 4 h 23 min down time per year
• Telecom
• 99.999% = 5 min
• Availability Zone, fully (?) independent computing systems
• Using two EC2 Availability Zones
%9999.99)1(11 22APP FP
© Y Raivio
9/14/201114
Everything as a Service
SaaS (Software as a Service)
– Ready to deploy application
– Salesforce, Gmail, SMS, voice
PaaS (Platform as a Service)
– No system administration
– Simplified development
– Scaling is provided by the PaaS framework
– Google Apps Engine, Microsoft Azure, Force.com
IaaS (Infrastructure as a Service)
– Computers owned by the cloud provider
– No hardware management issues
– Dynamic scaling of resources through virtualization
– Billing is calculated by usage only
– Amazon EC2
Sim
plicit
y
Evo
luti
on
To
tal
mark
et
40 B
€(2
011)
70
% S
aa
S&
Pa
aS
-3
0%
Ia
aS
© Y Raivio9/14/2011
Technologies
© Y Raivio
• Drivers
• Better utilization of HW (from 15% to 80%), saves energy and money
• Reduces system admin work
• Easier SW installation
• Hypervisors (VM manager): Xen, KVM (Kernel based VM), VMware
• Full (complete HW simulation), Para (interface between OS and HW) and HW-assisted virtualization
Virtualization
9/14/201116
APP
1
OS
CPU
APP
N
CPU
2
Source: Z. Ou, Virtualization Technology, T-110.7100, Autumn 2010.
..
OS
CPU
APP APP
1
OS
APP
N..
CPU
1
Hypervisor
Virtual Machines
..
Single task Multi task Hyper threading Virtualization
APP
1
OS-1
APP
N.. APP
1
OS-N
APP
N..
© Y Raivio
• No SQL, Not only SQL
• A family of databases for implementing (large) storage in the cloud
• Google BigTable, Amazon SimpleDB, LinkedIn Voldemort
• Good for web-scale data and analytics, not so great for transaction processing
• Data model not relational, rather a key-value store
• Scalable by nature
• ACID (Atomicity, Consistency, Isolation, Durability) relaxed in favor of BASE (Basically Available, Soft state, Eventually consistent)
• Challenges: interoperability, lock-in problem, data lifecycle, small data
Storage
9/14/201117
Source: R. Paivarinta and Y. Raivio, ”Performance Evaluation of NoSQL Cloud Database
in a Telecom Environment”, Closer 2011.
© Y Raivio
9/14/201118
Telecommunication Application
Transaction Processing (TATP)• Originally developed in
2003 to test HLRs based on SQL databases
• Simulates load on HLR database
• Ported for HBase, four tables denormalized into one adding redundancy
• 80% reads, 20% writes
Transaction
name
Type % Tables
Get-Subscriber-
Data
Read 35 Subscriber
Get-New-
Destination
Read 10 Special Facility,
Call Forwarding
Get-Access-
Data
Read 35 Access Info
Update-
Subscriber-Data
Write 2 Subscriber,
Special Facility
Update-Location Write 14 Subscriber
Insert-Call-
Forwarding
Write 2 Call Forwarding
Delete-Call-
Forwarding
Write 2 Call ForwardingHLR
MSC
Client N
MSC
Client 2
MSC
Client 1
…
© Y Raivio
9/14/201119
Measurement results
© Y Raivio
Service Level Agreement (SLA)
9/14/201120
Source: M. Murphy, ”Telco Clouds” [presentation], Cloud Asia 2010
• Research topics:
• Availability
• Latency
• Throughput
• Availability alone not enough
• Telecom users require more specific SLAs
• Sustainability?
• Penalties from violation?
• Monitoring tools important SLA Carrier grade 6 EC2 Large VMs
Availability 99.999 %99.95 % one zone
99.9999 % two zones
Latency < 150 ms < 50 ms (EU zone)
Throughput > 1000 msg/s >1000 msg/s
© Y Raivio
Measurement results
9/14/201121
Source: R. Paivarinta and Y. Raivio, ”Performance Evaluation of NoSQL Cloud Database
in a Telecom Environment”, Closer 2011.
© Y Raivio
Programming models
9/14/201122
• High data volumes
• Facebook: 1 PB, 2-3 TB added every day
• Web: 100 billion web pages -> 400-500 TB compressed
(duplicated across several clusters)
• eBay: 6.5 PB, 50 TB added every day
• Bottleneck in data transfer speeds (reads/writes from/to disks)
• Unlike disk capacity, bandwidth improves linearly
• Solution: read parallel from multiple disks
• How to change sequential programming to parallel
• Fault tolerance, Automated program partitioning
T = 1012
P = 1015
© Y Raivio
MapReduce
9/14/201123
• Google introduced MapReduce 2003/2004
• Hides from a programmer complexity of parallelization, fault-tolerance, data distribution and load-balancing
• Principle
• Iterate over a large number of records
• Extract something of interest from each (MAP)
• Shuffle and sort intermediate results
• Aggregate intermediate results (REDUCE)
• Generate final output
Source: Denis Shestakov, Cloud Computing seminar 24.9.2010
© Y Raivio
Traditional IT vs. Iaas, PaaS, SaaS
9/14/201124
Source: Rolf Harms and Michael Yamartino: The Economics of the Cloud, Nov. 2010.
© Y Raivio
• Bandwidth needed for VM migration
• Delay, jitter
• Location independent
Communications
9/14/201125
Source: Ambrust et al, ”Above the Clouds: A Berkeley View of Cloud Computing”, 2009
Bandwidth bottleneck: 1 TB drive, 1 Gbit/s I/O = 2 h 13 min
© Y Raivio
• Static
• Challenges
• Dynamics
• Machine learning
• Lack of APIs for monitoring and control
• Go-scheduling (computation close to data,
simultaneous computation)
• Analytics tools a hot topic
Provisioning and monitoring
9/14/201126
© Y Raivio
• Power Usage Efficiency (PUE) defined as:
• Typical PUE 2-3, state of the art 1.7
• 15% of all costs
• 59% IT equipment
• 33% cooling
• 8% power loss
• Power off idle machines
• Raise temperature
• Cold location, cheap energy (for example Finland)
• More fine-grain accounting
• Better algorithms
• E2E model required including public clouds
Power management
9/14/201127
tPowerITEquipmen
ityPowerTotalFacilPUE
Source: A. Greenberg, J. Hamilton, D.A. Maltz abd P. Patel, The Cost of a Cloud: Research
Problems in Data Center Networks, ACM SIGCOMM Computer Comm. Review, Jan 2009.
© Y Raivio
Privacy, Security and Trust
9/14/201128
Threat Description Target
Abuse and Nefarious Use of Cloud
Computing
Abusing anonymity for
spamming, malicous code
launch etc.
IaaS,
PaaS
Insecure Interfaces and APIs Anonymous logging, miuse of
resources
All
Malicious Insiders Misuse of internal information All
Shared Technology Issues Misuse of computation
resources
IaaS
Data Loss or Leakage Misuse of data All
Account or Service Hijacking Eavesdrop business All
Unknown Risk Profile Lack of control All
Source: CSA, Top Threats to Cloud Computing V1.0, March 2010
Other topics: regulation, privacy, service reputation management
© Y Raivio
9/14/201129
Public
cloud
Private
cloud
Telecom Cloud
SaaS
PaaS
IaaS
Support Systems
(MVNO/BSS)
Service Delivery
(SMSC)
Storage (HBase)
Computation (HLR)
Communication
Open Telco
SaaS
PaaS
IaaS
SaaS
PaaS
IaaSHybrid
Cloud
Eucalyptus
OpenStack
OpenNebula
Mobile Cloud Computing framework
Amazon EC2
End users
Adhoc
Cloud Mobile
Offloading
© Y Raivio
• Mobile access to cloud
• Using mobile context to enhance cloud based services
• Adhoc mobile cloud
• Mobile offloading: moving computation or/and data to cloud
• Already third of Facebook accesses from mobile
Mobile client
9/14/201130
© Y Raivio
Mobile offloading
9/14/201131
Source: B.-G. Chun and P. Maniatis, ”Augmented
Smartphone Applications Through Clone Cloud
Execution”, HotOS 2009.
Source: Kumar & Lu, ”Cloud Computing for Mobile Users:
Can Offloading Computation Save Energy ”, 2010
© Y Raivio9/14/2011
Conclusions
© Y Raivio
Challenges and opportunities
9/14/201133
Source: Ambrust et al, ”Above the Clouds: A Berkeley View of Cloud Computing”, 2009
# Challenge Opportunity
1 Availability of Service Use Multiple Cloud Providers
2 Data Lock-In API standardization
3 Data Confidentiality and Auditability Deploy Encryption, VLANs, and Firewalls
4 Data Transfer Bottlenecks Higher Bandwidth LAN Switches
5 Performance Unpredictability Flash Memory
6 Scalable Storage Invent Scalable Store
7 Bugs in Large-Scale Distributed
Systems
Invent Debugger that relies on Distributed
VMs
8 Scaling Quickly Invent Auto-Scaler
9 Reputation Fate Sharing Offer reputation-guarding services
10 Software Licensing Pay-for-use licenses
© Y Raivio
9/14/201134
•Remote and shared computing over the Internet
•Consists of components that communicate through APIs
!
•Simple architecture
•Efficient usage of CPU (>50%)
•Scalability
•Load balancing
•Low capex
•High availability
?
•Security & Privacy
•High usage of certain CPUs
•Interoperability
•Vendor lock-in
•High opex
•SLA critical
Pros and Cons
© Y Raivio9/14/2011
Questions?
Contacts:
Teacher: yrjo.raivio(at)aalto.fi, A122
Assistants: ramasivakarthik.mallavarapu(at)aalto.fi, PlayRoom
koushik.annapureddy(at)aalto.fi, A118
Course staff: t-110.5121(at)tkk.fi
© Y Raivio
1. Virtual Keynote by Doug Terry: ”Technology in the Cloud – Plus some
Challenges and Opportunities”, Microsoft Research, June 1, 2011, available
from http://techpack.acm.org/cloud/
2. Armbrust, Michael, Fox, Armando, Griffith, Rean, Joseph, Anthony D., Above
the Clouds: A Berkeley View of Cloud Computing, Feb. 10, 2009. Available at:
http://www.eecs.berkeley.edu/Pubs/TechRpts/2009/EECS-2009-28.pdf
3. Cloud Software program, D1.1.1 Technical Report: Cloud Computing
Technologies, June 27, 2010, available at:
http://www.cloudsoftwareprogram.org/results/i/26999/1569/d1-1-1-technical-
report-cloud-computing-technologies
4. Lee Badger, Tim Grance, Robert Patt-Corner and Jeff Voas: Draft Cloud
Computing Synopsis and Recommendations, Recommendations of the
National Institute of Standards and Technology, May 2011, available at:
http://csrc.nist.gov/publications/drafts/800-146/Draft-NIST-SP800-146.pdf
5. Rolf Harms and Michael Yamartino: The Economics of the Cloud, Nov. 2010,
available at: http://www.microsoft.com/presspass/presskits/cloud/docs/The-
Economics-of-the-Cloud.pdf
Reading material
9/14/201136