14. Definition:
Clouds are hardware-based services
offering compute, network and
storage capacity where:
Hardware management is highly
abstracted from the buyer
Buyers incur infrastructure costs as
variable OPEX
Infrastructure capacity is highly
elastic (up or down)
McKinsey & Company
18. By a 5‐to‐1 ratio,
companies trust internal
IT systems over cloud‐
based technologies due to
fear about security
threats and loss of control
of data and systems.
Avanade Inc.
19. They are right...
Streamload: On June 15,
2007 system
administrator's script
accidentally misidentified
and deleted quot;good dataquot;
along with the quot;dead dataquot; of
some 3.5 million former
user accounts and files.
20. Or are they?
LAX Los Angeles 1,200
MIA Miami 1,000
JFK New York 900
ORD Chicago 825
EWR Newark 750
0 300 600 900 1200
10 U.S. airports with the highest weekly frequency of laptop loss
Airport Insecurity: The Case of Missing & Lost Laptops, Ponemon Institute LLC
21. This get scary...
Did not protect sensitive
65%
information contained on laptop
57%
Worry about losing their laptop
Laptop contains
53%
confidential company information
42%
Data on laptop is not backed up
0 0.175 0.35 0.525 0.7
Airport Insecurity: The Case of Missing & Lost Laptops, Ponemon Institute LLC
22. Most likely causes of data
breach?
Negligent insiders 75%
Outsourced data 42%
Malicious insiders 26%
Social engineering 2%
Hackers 1%
0 0.2 0.4 0.6 0.8
2008 Study on the Uncertainty of Data Breach Detection,Ponemon Institute LLC
23. IT environment where data
breaches occur
Off-network devices 58%
Networks 50%
Mainframes 41%
Paper files 39%
Backups 18%
0 0.15 0.3 0.45 0.6
2008 Study on the Uncertainty of Data Breach Detection,Ponemon Institute LLC
24. Ability to detect the loss or
theft of confidential
information
31%
25%
18%
16%
10%
Somehow
Very confident Confident confident Not confident Unsure
2008 Study on the Uncertainty of Data Breach Detection,Ponemon Institute LLC
33. ACID -> BASE
Traditional approaches don’t
scale
BASE - basically available, soft
state, eventually consistent:
BigTable, SimpleDB,
Cassandra, Dynamo
34. GoodData: Innovate
vs. leverage?
Processing Power
Cloud makes ROLAP approach possible
Elastic Scale
IT builds for peak load, we don’t have to
Multi-Tenancy
Single instance across 1000s of customers
Stateless
Massive load balancing (shared nothing)
36. Public cloud classes
AWS MS Azure Google AE
Predefined
CPU x86 .Net
framework
EBS, S3, SQL, Azure
Storage SimpleDB store
BigTable
Network Declarative Automatic Fixed
37. Cloud APIs
True SOA
Loosely coupled - REST, Atom
Encapsulate cloud services:
Control APIs
Data APIs
Application Functionality APIs
40. If you want to
change the
game, change
the economics
of how the
game is
played
Alan M. Webber
41. Startups in the cloud
Infrastructure labor savings
No CAPEX: Less equity goes to VCs
Unpredictable demand (up and down)
42. Succeed (or fail)
faster
$500k to start technology company
Big bets aren't as big anymore
Easier for startups to adapt to shifts
Level playing field for startups
43. IT vs. Clouds
Losing their monopoly on the
infrastructure
It’s all about economics
45. Fixed pricing
Most widely used pricing
No supply/demand
Simple, predictable
AWS, Google App Engine:
CPU, Storage, network traffic
46. Variable pricing
Reserved instance price (AWS):
“I have 10 instances running 24x7”
Spot price/future price:
“I want 1,000 instances at the end of
the quarter”
Off-peak pricing:
“Run my MapReduce app 10 hours
every day”
49. Private clouds
Violate #2 of our cloud definition:
Buyers incur infrastructure costs
as CAPEX
Virtualization on top of traditional
enterprise IT stack
Encapsulation of IT infrastructure
Scale?
50. Economies of scale
Technology Medium-sized DC Very Large DC Ratio
$95 per Mbit/sec/ $13 per Mbit/sec/
Network 7.1
month month
$2.20 per GByte / $0.40 per GByte /
Storage 5.7
month month
140 Servers / >1000 Servers /
Administration 7.1
Administrator Administrator
HAMILTON, J. Internet-Scale Service Efficiency. In Large-Scale Distributed Systems and Middleware (LADIS) Workshop (September 2008)