CloudOps Summit 2012, Frankfurt, 20.9.2012 Track 2 - Build and Run
by Nigel Sanctuary, VP Propositions at Kognitio (www.kognitio.com)
http://cloudops.de/sprecher/#nigelsanctuary
Find the video of this talk at http://youtu.be/wQrHQNOMlKc
2. Question
JUST THE
IS THE CLOUD LATEST
BUSINESS
COUTURE?
BI Process
outsourcing
3. Answer: Well, Yes…..But…
ASP model of the 1990’s
Established
Managed Services
Business Process Outsourcing
SaaS
Newish/ig
Virtualisations
Flexibility
4. The DB traditional performance management
model
High
Cost + Capacity Complexity and demand of
(Demand & Supply) Business intelligence
Growing Data Volumes
Develop & Manage + Capacity = Cost
Cost of Data
Environment
System Performance
Low
Time
5. Cost and risk aligned with need and budget
High
Cost + Capacity Complexity and demand of
Growing Data
Volumes
(Demand & Supply) Business intelligence
Low Cost Data Environment Increments
Performance aligned to
incremental growth
Low
Time
6. If these are the drivers, what must we check?
CIO Research Center
7. Let’s start with a definition
Cloud Computing: Cloud computing is a model for enabling openly
available, convenient, on-demand network access to a shared pool of
configurable computing resources (e.g., networks, servers, storage,
applications, and services) that can be rapidly provisioned and released with
minimal management effort or service provider interaction.
U.S.
Government's National Institute of Standards and Technology
9. Cloud Service Types
Software as a Service (SaaS): Application only – Salesforce.com
10. • Platform as a Service (PaaS): Applications provided on a
preconfigured hardware and configured by customer –
Kognitio, Engineyard, SAAVIS
Database Systems
Portal Products Testing
Business Process
Management
PaaS Backup &
Recovery
State
Integration Network management
tools Management
11. • Infrastructure as a Service (IaaS): The consumer uses
"fundamental computing resources" such as processing
power, storage, networking components. AWS, SAVVIS
Switching
Servers Virtualisation
IaaS
Virtualisation
Networks
12. My Scope
Lot’s
of e
xper
ienc
e
Database Systems
Portal Products Testing
Business Process
Management
PaaS Backup &
Recovery
State
Integration Network management
tools Management
14. Kognitio’s Cloud Service
Customer
Webs Site
BI App.
Volume Location Time
store
Private Cloud Public Cloud
15. What the client’s buy : High performance analytical
platforms
Data moved into Many CPU cores Queries run directly
high speed memory Shared-nothing against data in memory
Very high performance
Handles any analytical workload
CPUs
Server 1
RAM merged CPUs
Server 2 together
across
servers into a
shared fabric CPUs
Server 3
Server 4 CPUs
16. Made possible by the collapse of in Memory
costs
RAM $/TByte (10 Years)
$400.000
$367.002
$350.000
$300.000
Costs have/are falling
$250.000
$200.000 $202.375
RAM $/TByte
$150.000 $156.238
$135.266
$100.000
$85.983
$76.546
$50.000
$40.894
$22.020
$-
2002 2003 2004 2005 2006 2007
$18.874
2008 2009 2010 2011 2012
$5.138
Source: Feb, 2012, John C. McCallum
17. The Check List: For Database services
• What do you NEED to achieve
• What are the benefits you want from Cloud v in house
– In house V outsource
– Pricing Options
– Contract terms v price
• Security Levels:
– “Collected data should be kept secure from abuse” EE Data Protection Directive
– Managing Security – Buying services/responsibility
– How can you put the data in the cloud (Hipaa)
• Availability
– Volume
– Flexibility - Scaling
– User Base and workload
– Service Levels/ Responsibility
– Application integration/System Integration
18. If these are the drivers, what must we check?
CIO Research Center
19. The question: Which component
Operational database Computer/ Automated
Telephony Call Delivery
Integration
Branch
Operational
Reporting Interactive
Transactional Voice Response
Database
Correspondent
Product Agent
systems Platform/
Operational
Operational Data staging Co-browsing
Data Store PSTN
Transactional
Web Content
Other op’s
Systems
database
Risk
Management/ Decision Contact
Business Decisions Engine Cookies
Intelligence centre
Reporting Tags
Analysis Campaign
Data
Application Management
Analytical Data Delivery Environment
Scores
Platform Marketing
Analysis/ Telephone
Decisions Campaign
outbound campaigns
Customer/event flags
Campaign Mail
Analytical Database
Campaign Data
20. What do you need it for
Application Processing
• Common solutions in IaaS,
SaaS
• High transactions volumes
• OLTP
• Need response, flexibility,
management, capacity
• Dynamically expanding and
Real Time Reporting contracting user community
• Common solutions in PaaS,
SaaS (Microstrategy
• High transactions volumes
• SMP/MPP
• Processing dependent,
management, capacity
• Large User Community Analytical batch Processing
• Solved through PaaS, SaaS
• Mid/Large batch data volumes
• Off line analytical processing
(MPP/SMP)
• Need speed, complexity
management, capacity
• Smaller user community
21. BIG DATA – More in it
Transactions
Business
Operations
Contact
22. What is the anticipated workload
Campaign Self learning
Behaviour Management algorithms
Analytical Complexity
modelling
Statistical
Analysis Dynamic
Content
Process
Reporting On-Line Fraud
detection
Clustering
Technology/Automation
Analytical Platform as a Service
23. What is the User Community
Impact on process capacity
• Demand Spikes
• Complex Queries
• Scheduled Capacity
• Volume of users
24. Cost and Trade Off
• What are the benefits you want from Cloud v in house
– In house V outsource
– Pricing Options
– Contract terms v price
25. Pricing Option Trade Off’s
In-House v Outsource-
1
=
No
Cost
10=High
Cost
In
House Outsource Comment
SLA
Control 5 5 Premiums
for
higher
SLA's
in
Outsource,
Shared
Service
cost
Capex
Budget 10 3 Outsource
No
Capex
options
available
Flexibile
per
usage
models
allow
budget
control.
SAAVIS
and
Budget
Control 8 6 AWS
both
offer
good
but
different
solutions
In
House
tend
to
have
faster,
focused
performance
control,
Performance
Control 4 6 Private
Cloud
is
best
if
this
is
a
priority
Shared
resources
and
system
performance
management
is
a
System
Management
Costs 8 6 key
benefit
of
cloud
System
Maintenance
responsibility 10 4 The
key
advantage
of
Outsource
models
PaaS
Outsource
include
charges
for
maintenance
and
Software
Maintenance 10 7 upgrade
using
shared
resources
Integrated
architectures
are
pre
constructed
requirinf
only
Implementation
Costs 10 3 confirguration
in
the
outsource
model
Total
Cost
Score 65 40 About
38%
improvement
if
you
optimised
the
benefits
27. Cost savings
• Considerations:
– Long term v Short term costs in the cloud
– Trade off of savings v functionality
– Total Cost of Ownership – Management, Skills, Hardware,
Applications, Networks
– Create you measure of success
– Target your improvement goal and buy components and
services to support it
28. Security
• Security Levels:
– “Collected data should be kept secure from abuse” EE
Data Protection Directive
– Managing Security – Buying services/responsibility
– How can you put the data in the cloud (Hipaa)
30. The fear
Unauthorised access
Data theft
Rogue user abuse
Data in transit
Physical Security
31. Investment in Cloud Security
Tier 3 dataAccess Control
VPN/SSH centres
Encryption
Monitoring Systems
Some good examples
http://aws.amazon.com/security/
http://www.rackspace.co.uk/managed-hosting/business-need/
security-compliance/
32. Security service options
Client
monitoring
File RootAdmin SysAdmin
Encryption Monitoring SSH/FixedIP
Data CPUs
X
Deploymentmin
Server Farm
Data RAM merged CPUs
together
across
servers into a
Data shared fabric CPUs
Data CPUs
VPN/ SysAdmin LinuxAdmin
PGP User Privileges Manage S’ware
33. Planning for use
• Availability
– Volume
– Flexibility - Scaling
– User Base and workload
– Service Levels/ Responsibility
– Application integration/System Integration
35. Manage Analytical Platform System to Demand
Persistent
Data Store RAM Servers
• Store Increasing
Data
• Flexible, granular
Data views of any data
Source
• Use RAM capacity
Data according to
demand
• Drop and Pin Data
Data in RAM
• Valuable RAM is
Data controllable
• Cheap Disk allows
Data Views/ you to lose nothing
Queries
SSD/Disk
37. Public Cloud
Can you create
persistent data layers
independent from
processing?
Can you switch data
between operational
schemas within one DB
installation?
Can you flex hardware
scale to scheduled
demand patterns?
38. Manage and re-configure for capacity
Kognitio Appliance: Industry Standard Hardware
Applicatio
n Source
data
Single Capacity, Multi-Use Kognitio Licence Analytics
Analysis Single DB instance Single DB instance Application
Single DB instance
Product & Reporting
Service Sandpit 1 Application
Contact History Development
Catalogue
& Quota
Loyalty Test & Geo Location
Management Management App
Development
Single DB instance
Geo Location
Info Service Single DB instance Development/
Permissions Sandpit 2
Database Test interface
Inventory
Management
Dynamic platform
Repository
ETL 1
Device management
Test
Test
Catalogue
Applications
Offer and ETL 2
Assets Data Analytics Database
Repository
System
management &
B2 Customer Data Staging for ETL 3
data Store Import/ Export
Capacity
Management Export to
application
data
39. Private/Hybrid Cloud
• Partition the system
to support differing
workloads within one
DB installation
• Support Mass
reporting and data
mining from one
platform
40. A sample solution
Data
Service & Access requirements
Kognitio Service Support
Client/ Prospect
ASP DB SFTP/other media Data Load
Fixed IP
Team
Report/
Query Access /Service Config. Management
Requests Client Direct Interface
Firewall
Client Results
Internet Interface
Internet
Direct
Interface
Client 1 Client 3
Kognitio Cloud app store Source data
Tools Client 2 Client 4
Application Server
Cloud Servers (Public or private)
41. The Check List: For Database services
• What do you NEED to achieve
• What are the benefits you want from Cloud v in house
– In house V outsource
– Pricing Options
– Contract terms v price
• Security Levels:
– “Collected data should be kept secure from abuse” EE Data Protection Directive
– Managing Security – Buying services/responsibility
– How can you put the data in the cloud (Hipaa)
• Availability
– Volume
– Flexibility - Scaling
– User Base and workload
– Service Levels/ Responsibility
– Application integration/System Integration