Mais conteúdo relacionado Semelhante a Dynamic Software Defined Network Infrastructure Test Bed at Marist College (20) Dynamic Software Defined Network Infrastructure Test Bed at Marist College1. Dynamic Software Defined Network (SDN)
infrastructure test bed at Marist College with IBM
and ADVA Optical Networking.
Todd Bundy
Director Global Alliances-Enterprise ASM
ADVA Optical Networking
tbundy@advaoptical.com
203-546-8230
2. © 2013 ADVA Optical Networking. All rights reserved. Confidential.22
Our speaker today
Todd Bundy
Business Development Director,
ADVA Optical Networking
tbundy@advaoptical.com
Special Thanks:
Robert M. Cannistra
School of Computer Sciences and
Mathematics
Marist College
robert.cannista@marist.edu
Special Thanks:
Casimer DeCusatis
Distinguished Engineer,
IBM STG –eSystems Dev Lab
decusat@us.ibm.com
3. © 2013 ADVA Optical Networking. All rights reserved. Confidential.33
Michael Haley,
Distinguished Engineer, IBM CHQ – El Cloud
Infrastructure
haleym@us.ibm.com
Jim Theodoras ,
Senior Director Technical
Marketing, Marketing & Strategy,
ADVA Optical Networking
JTheodoras@advaoptical.com
Christian Illmer,
Business Development Senior Director
ADVA Optical Networking
cillmer@advaoptical.com
Special Thanks
4. © 2013 ADVA Optical Networking. All rights reserved. Confidential.44
The Need for SDN
5. © 2012 IBM Corporation5
More Complex / Yet More Consumable
Data and data management
Discovering insights
Interaction
Fast Pace
Evolving business eco-system
Dynamic scalability
Minimize time to value & update
Contextual Overload
Proliferation of sensors and devices
Demand for personalization
Just in time
IT Mega Trends
Technology Drivers
Mobile - Social - Cloud -
Big Data / Analytics
Growing Scale/Lower Barrier of Entry
Users
Transactions, Computations
Data, Bandwidth
6. © 2012 IBM Corporation6
Smarter Commerce Example
Service
Buy
Sell
Market
Superbowl
add
e Big Data Analytics
detect social
media spike
Supply chain
updated
Target
marketing
Web
transactions
Update for
brick & mortar
In-store
transactions
Client
follow-up
Channel
follow-up
45% of customer seek advice from others, before a purchase.
53% of folks on twitter recommend companies and products.50% of customer value is in the hands of suppliers.
90% of marketers say web data is important for analytics & decisions.
85% of clients expect seamless experience across channels.
86% of clients stop doing business with a company
due to bad experience.
60% of clients are willing to pay more for a better experience
7. © 2012 IBM Corporation7
Smarter Commerce Example
Service
Buy
Sell
Market
Superbowl
add
e
Big Data
Analytics
detect social
media spike
Supply chain
updated
Target
marketing
Web
transactions
Update for
brick & mortar
In-store transactions
Client
follow-up
Channel
follow-up
Physical
Fabric
Network
Hypervisor
Intrusion
Prevention
Firewall
Web Servers Application
Server
Firewall
Load
Balancer
Database
Cluster
Clients require a fast way to respond
to their market demand
• Today it takes days. For example, at
OFC 2013 Cisco’s SVP of IT, John
Manville stated it takes Cisco’s IT 5
days to fully bring a multi-tier
workload on-line, includes:
configuring network appliances,
storage,etc…
• For this example to be possible, 5 days
needs to be 5 minutes or less.
8. © 2012 IBM Corporation8
Big Data: Fueling Smarter Commerce Cycle
Video
Text
Exa
Peta
Tera
Giga
DataVolume
1990’s 2020’s2000’s 2010’s
Structured data
Audio
Image
Med
High
Low
ComputationalNeeds
SophisticationofAnalysis
Digital Marketing
10+% of video views
Wide Area Imagery
100’s TB per day72 video hrs/minute
Media
Source: IBM Market Insights based on composite sources
Safety / Security
Healthcare
Customer
1B camera phones
1B medical images/yr
10s millions cameras
Enterprise Video
Used by 1/3 of enterprises
9. © 2013 IBM Corporation9
System Networking Requirements
Automate connectivity of
multi-tier system patterns
(Lower OpEx)
Hyperscale
(Lower CapEx)
De-couple virtual network
from physical network
(Lower OpEx)
Flow optimization
(Lower CapEx)
Provide global network visibility
with “real-time” control
(Lower CapEx)
New network interaction
paradigm (Lower OpEx)
Tenant 1 cloud
Network Hypervisor
Tenant 2 cloud
Site A Site B
Storage
Firewall
Web
Server
Database
Application
Server
Intrusion
Prevention
Hours
to Days
Minutes
10. © 2012 IBM Corporation10
Cloud needs elasticity and multi-tenancy
Big Data & Analytics needs Application controlled, optimized fabrics
Mobility needs guaranteed end-to-end SLA
DC Virtualization needs agility of a fully virtualized environment
Why is the Network usually a boat anchor in today’s data center?
Today’s Data Center Network
often gets in the way of client trends
Today’s protocols make multi-tenancy complex.
Box-level services & discrete physical infrastructure prevent true elasticity.
Network virtualization is lagging other parts of DC Infrastructure.
Complexities of physical network (box-type, SW version;
capabilities; vendor inconsistencies) drag agility.
Box-by-box protocols and behavior prevent anything end-to-end
and results in error-prone security and QoS
Closed networks cannot be programmed via business applications,
resulting in under-utilized networks.
But….
But….
But….
But….
11. © 2013 IBM Corporation11
Pods of IT Resources
Embedded Blade, Virtual Switches
Overlays Virtualize L2/3 functionsWAN
SAN
Pools of
Virtual Appliances
stacked switches
lossless Ethernet
FCoE, TRILL, & more
Virtualized &
SDN Enabled
FCoE
Gateway
FCoE Storage
SDN controller
ODIN – The Open Datacenter Interoperable Network
An industry standard point of view on SDN & network virtualization
http://www-03.ibm.com/systems/networking/solutions/odin.html
ONF
The Linux Foundation (nonprofit
organization dedicated to open
source development) has
announced the OpenDaylight
Project (www.opendaylight.org )
OpenDaylight SDN Project
Includes OpenFlow protocols, an open
controller, virtual overlay networks
(including DOVE), protocol plug-
ins, and switch device
enhancements
12. © 2013 ADVA Optical Networking. All rights reserved. Confidential.1212
Virtualization and Beyond
13. © 2013 ADVA Optical Networking. All rights reserved. Confidential.1313
What is SDN?
• Today, routers control themselves, passing messages
between them
• SDN separates the Control and Data planes
• A separate, centralized controller then takes command
Control
Data
Data Data
Data
Data
Control
Control
Data
Control
Data
Control
Data
14. © 2013 ADVA Optical Networking. All rights reserved. Confidential.1414
Do we want Distributed or Centralized?
14
• Ethernet topologies were built
distributed
Scalable but hard to monitor
• Openflow topologies (today) are
centralized
Control-data separation forces this model
Strengths of one approach are weaknesses of the other
Centralized is better suited for modern cloud applications
IBM System Networking
15. © 2013 ADVA Optical Networking. All rights reserved. Confidential.1515
Better topologies
• Multi-tiered tree topologies
• High oversubscription
• Expensive, high bandwidth uplinks
• Robustness of higher tier product
has been a concern
• 2-tiered mesh or Clos topologies
• Oversubscription only to WAN/core
• Large cross sectional bandwidth
(TOR bandwidth is cheap)
• Mature Layer 2/3 softwareIBM System Networking
16. © 2013 ADVA Optical Networking. All rights reserved. Confidential.1616
SDN in a Single Sentence
• The hope/expectation of better efficiency/utilization through
identifying flows of packets and making smarter holistic decisions
• A good analogy would be the Air Traffic Control system
*
flightradar24.co
m
*
http://virtualskies.arc.nasa.gov/atm/6.
html
* http://mattsflight.blogspot.com
17. © 2013 ADVA Optical Networking. All rights reserved. Confidential.1717
IBM System Networking
17
Packet or Flow Switched?
• Ethernet topologies are packet
switched
Statistical link utilization
• Openflow topologies (today) are
flow switched
Application level network control
18. © 2013 ADVA Optical Networking. All rights reserved. Confidential.1818
SDN Model: Applications influence the network
18
VM
VM
VM
VM
App App
Network
Hypervisor
Program
& Instruct
Old Model
IBM System Networking
19. © 2013 ADVA Optical Networking. All rights reserved. Confidential.1919
How do you extend SDN to the Metro?
Data Data
Data
Data
SDN Control
Data Center C
Data Data
Data
Data
SDN Control
Data Center A
Data Data
Data
Data
SDN Control
Data Center B
150km
75km 50km
…but what about:
• Billing
• Carrier interworking
• Stacking of SDN controllers
• Oversubscription, lack of resources
20. © 2013 ADVA Optical Networking. All rights reserved. Confidential.2020
Metro connectivity:
• Data center distances < 100km
• Ethernet, Fiber Channel, some Infiniband
• DWDM w/ minimum processing (latency)
Optical Network Virtualization
Core connectivity:
• Globally distributed data centers
• IP/MPLS over OTN/DWDM
• Standard core network technologies
Metro
Core
Increased resource efficiency and reduced interconnection costs
by optical capacity sharing.
Programmable optical
connectivity for
cloud data center
operators & tenants
21. © 2013 ADVA Optical Networking. All rights reserved. Confidential.2121
Virtualization Beyond DC Boundaries
Web 2.0 data centers Global carriers
From “warehouse-scale computers” to “the network is the computer”.
22. © 2013 ADVA Optical Networking. All rights reserved. Confidential.2222
Optical transport and SDN
• Decades of work have yielded today’s agile core networks
• Unfortunately, the information to make intelligent decisions resides
at higher layers
• Problem is made worse by today’s flow dominated traffic
Router Router
Intelligent
MUX
Hybrid
EDFA/RAMAN
Amp
Agile
Core
Network
Gridless
ROADM
Coherent
Receiver
23. © 2013 ADVA Optical Networking. All rights reserved. Confidential.2323
Optical OpenFlow Extensions
C
Any λ
λ Block
WSS
WSS
WSS
WSS
λ1λ2λ3
TX
WSS
1:NWSS
1:N WSS
1:NWSS
1:N WSS
1:N WSS
Connectivity & Topology Discovery
Signal Mapping & Format Compatibility
Optical Node Model:
Wavelength Continuity &
Dynamic Switching Constraints
Optical Performance Constraints
Sequential Lightpath Setup/Teardown
Optical Power Balancing
OpenFlow Support for
FSP 3000 ROADM platform
Optical extensions required to deal with the analogue nature of the optical layer.
OSC
(Out-of-band)
24. © 2013 ADVA Optical Networking. All rights reserved. Confidential.2424
Optical transport and SDN
Direct vs Indirect Model
• How to best extend SDN to transport layer?
• Direct control yields most potential benefits at cost of complexity and
latency
• Indirect control is easier to implement, and provides a migration path
IndirectDirect
25. © 2013 ADVA Optical Networking. All rights reserved. Confidential.2525
Dynamic Infrastructure Test Bed
26. © 2013 ADVA Optical Networking. All rights reserved. Confidential.2626
Marist SDN Lab Objectives
• Initially funded by an IBM STG University Alliance Team
Faculty Grant received in 4th quarter 2011
• Compliment IBM software-defined networking research &
development
• Establish a pipeline of skilled students
• As of December 2012, we are proud to have successfully
completed all of the initial objectives set forth and we are
moving full steam ahead in creating new ones!
27. © 2013 ADVA Optical Networking. All rights reserved. Confidential.2727
NY State Center of Excellence for
Cloud Computing and Analytics
• $3M grant
• Awarded Dec. 2012
• Student Research
• Business Intelligence Cloud
• Incubation Center
• Workforce Development Center
28. © 2013 ADVA Optical Networking. All rights reserved. Confidential.2828
MARIST: Dynamic Infrastructure Test Bed
VM Cluster
ADVA FSP
3000 Site A
Site C Site B
ADVA
FSP 3000
ADVA
FSP 3000
StorageStorage
IBM V7000
Storage
dual
10G
dual
10G
dual
10G
dual
10G
IBM G8264
OF Switch
IBM G8264
OF Switch
Floodlight
Controller
OpenDayLight
Controller
29. © 2013 ADVA Optical Networking. All rights reserved. Confidential.2929
MARIST: Adding System Z
VM Cluster
ADVA FSP
3000 Site A
Site C Site B
ADVA
FSP 3000
ADVA
FSP 3000
Storage
Storage
IBM V7000
Storage
dual
10G
dual
10G
dual
10G
dual
10G
IBM G8264
OF Switch
IBM G8264
OF Switch
OpenDayLight
Controller
Floodlight
Controller
30. © 2013 ADVA Optical Networking. All rights reserved. Confidential.3030
VM Cluster
Dynamic Infrastructure Test Bed
ADVA FSP 3000
Site A
Site C Site B
ADVA
FSP 3000
ADVA
FSP 3000
StorageStorage
IBM V7000
Storage
dual 10G dual 10G
dual 10G dual 10G
IBM G8264
OF Switch
OpenFlow
OpenFlow Controller (VM)
• Floodlight
• IBM Controller
• OpenDaylight
ADVA OpenFlow Agent (VM)
• OpenFlow v1 northbound
• ADVA control plane southb.
31. © 2013 ADVA Optical Networking. All rights reserved. Confidential.3131
Dynamic resource orchestration
Control plane view
OF Controller
(ie: OpenDayLight Controller)
ADVA OF Agent
ADVA FSP 3000
OF Switch OF Switch OF Switch
WAN Orchestration
Web I/F
Fully automated
Virtual Compute
Environement
STEP 1 STEP 2
Request/grant
Resources
32. © 2013 ADVA Optical Networking. All rights reserved. Confidential.3232
SDN for Dynamic SP Infrastructure
• Provisioning for peak traffic is losing battle, and only getting worse.
• Answer is dynamic network infrastructure.
Daytime Configuration
All Offices/Sites working
Nighttime Configuration
Backup between A/B
Double the bandwidth
Other Configurations
Site B to C
Site C to A
Site A
Site B
Site C
1x 10G
1x 10G
1x 10G
Site A
Site B
Site C
2x 10G
Site A
Site B
Site C
2x 10G
Site A
Site B
Site C
2x 10G
Site A
Site B
Site C
1x 10G
1x 10G
1x 10G
33. © 2013 ADVA Optical Networking. All rights reserved. Confidential.3333
Use Cases
Bandwidth calendaring Cloud bursting
Secure multi-tenancyWorkload balancing
Transactional nature of DC-to-DC traffic (bulk data transfers)
offers opportunities for optical bandwidth-on-demand.
Cloud DC
Private
Datacenters
Tenant 1
Tenant 2Load Load
34. © 2013 ADVA Optical Networking. All rights reserved. Confidential.3434
Summary
• Optical network virtualization offers cloud providers & tenants
high-bandwidth, low-latency connectivity on demand.
• Different models for optical network virtualization exist.
• A compromise between hiding the optical complexity and
exposing the optical topology is required.
• Open approaches based on standardized GMPLS or emerging
OpenFlow technologies are possible.
35. © 2013 ADVA Optical Networking. All rights reserved. Confidential.3535
Q & A
36. info@advaoptical.com
Thank You
IMPORTANT NOTICE
The content of this presentation is strictly confidential. ADVA Optical Networking is the exclusive owner or licensee of the
content, material, and information in this presentation. Any reproduction, publication or reprint, in whole or in part, is strictly
prohibited.
The information in this presentation may not be accurate, complete or up to date, and is provided without warranties or
representations of any kind, either express or implied. ADVA Optical Networking shall not be responsible for and disclaims any
liability for any loss or damages, including without limitation, direct, indirect, incidental, consequential and special damages,
alleged to have been caused by or in connection with using and/or relying on the information contained in this presentation.
Copyright © for the entire content of this presentation: ADVA Optical Networking.
Notas do Editor The past year has witnessed a remarkable change in the IT marketplace with many new entrants, a significant uptick in the number of mergers and acquisitions, growing investment by the venture capital community in enterprise focused companies, and an increasingly competitive and dynamic ecosystem. The technology drivers of social, mobile, cloud, Big Data, and analytics have led to the identification of four mega trends in the 2013 GTO.Growing Scale / Lower Barrier of Entry – Massive expansion in the number of users, sensors and smart devices, channels, relationships, transactions, computations, and the amount of structured and unstructured data is occurring. In 2013, the total of all digital data created is forecast to reach four zettabytes, or 4 x 1021, nearly four times 2010 volumes. A growing number of easy-to-consume application programming interfaces (APIs) for integration with applications, platforms, and infrastructure is lowering the barrier to entry. Increasing Complexity / Yet More Consumable – The volume, variety, velocity, and veracity of data is contributing to the increasing complexity of data management, workloads, and advanced analytics needed to discover insights. For example, novel algorithms to analyze large graphs are resource intensive and new kinds of data management for these graph structures will be needed in the future. Users are becoming more sophisticated and demanding interactive tools for business and scientific visual analytics. The mobile phone has evolved from a simple voice device to a multimedia communications tool capable of uploading and downloading data, text, audio, and video while also functioning as a global positioning system, wallet, FM radio, television, alarm clock, thermometer, address book, newspaper, camera, and more.Fast Pace – Disruptive new models for development and consumption are emerging and increasingly penetrating the enterprise ecosystem resulting in rapid innovation and decreased time to value. Componentized hardware and software is available off the cloud. New web front-end pay-per-use service models are experiencing rapid growth. Social media is influencing how developer and open source communities are engaging in this global ecosystem. Massive open online courses are experiencing exponential growth making education and training more accessible. Contextual Overload – The proliferation of sensors and devices and the explosive growth in structured and unstructured data are causing information and contextual overload. With the increasing affordability and sophistication of smart devices, new opportunities exist to provide contextually aware and personalized services based on user views, desires, preferences, and location delivered just-in-time. Initially funded by an IBM STG University Alliance Faculty Grant which was received in the fourth quarter of 2011.This joint study initiative compliments IBM’s internal research and development in the area of software-defined networking.The initial objectives for this project were to: - establish an SDN development lab with IBM G8264 OpenFlow capable switches - to contribute to the open source community by making contributions to the openflowdistro and open source community - to establish first use cases for OpenFlow - to share our knowledge by writing research papers and presenting at conferences - help students gain knowledge around SDN and OpenFlow - provide the students exposure to professional activities such as conferences and presenting at the IBM Academic of Technology - and to form partnerships with other organizations, companies and clients.Marist has been truly honored and blessed to work with IBM on our Joint Study initiatives. IBM first approached Marist in 1988 to help build a collaborative environment that would provide leadership, guidance, and support to the IBM Academic Initiative through course development, faculty training, and faculty research which allows faculty and students alike access to IBM hardware and software worldwide. This helps Marist and IBM move forward with advances in technology.One of the more recent joint study projects that we’ve been working on is our Software-Defined Networking Lab which was originally funded through a faculty award that I received from IBM which to-date is close to $200,000 dollars. And won’t it be nice when we demonstrate all of this running on anopenflow network. Basic Topology Overview Basic Topology Overview Basic Topology Overview