1. Alberto Leon-Garcia and Thomas Lin
E2E Orchestration on Multitier Clouds based on Software-Defined Infrastructure
Alberto Leon-Garcia and Thomas Lin
University of Toronto

2. Agenda 09:00 - 10:00 SAVI Introduction 10:00 - 10:15 Coffee Break
10: :45 SAVI Basics & Setting up command line interface
10: :15 Multi-Region Orchestration
11: :00 Layer 2 WAN & Service Chaining
LUNCH + OTHER DEMOS
15: :30 SAVI Monitoring Architecture (MonArch) & Video
15: :00 SAVI IoT Introduction
16: :00 Multi-Region Containerized IoT Platform

3. SAVI Introduction

4. Context: Smart Application Platforms
The Challenge
By 2050
Over 70% of world population will live in cities
Occupy 2% of landmass
Consume 75% of resources
The Opportunity
To enable livable and sustainable cities and urban regions
economic, environmental, social
Our Focus
Platforms to enable Smart City Applications
Applications Platforms for Smart Transportation
Converged Cloud computing, SDN, and IOT

5. Application Platform Requirements
Real-time Situational Awareness
Continuous Monitoring
Data Collection & Historical Record
Flexibile & Extensible data types
Learning and Intelligence
Analytics & Machine Learning
Standard & Custom Reporting

6. Application Platform Requirements
Visualization
System Dashboard Available on Demand—Anytime, Anywhere, Authorized
Trending & Forecasting
KPI Analysis
Smart Applications Enablement
APIs provide real-time, historical, value-add data
Emergency Response
Event Planning & Management
Traffic Control, Planning, Optimization

7. Demo: CVST Portal of Greater Toronto Area

8. More Change on the Way: Smart Power Grids

9. Sensor and Actuator Gossamer
IOT at Scale
Real-Time Situational Awareness
Learning and Intelligence
Visualization
Smart Applications Enablement ?
Sensor and Actuator Gossamer

10. Brief Introduction to Cloud Computing

11. Definition of Cloud Computing
Cloud computing is a new computational model that allows users to consume IT resources (computing, storage, software, network) through network as a utility -- just like electricity -- rather than having to build and maintain computing infrastructures
Storage
Server
Software
Everything as a Service

12. Key Technology for Cloud Computing: Virtualization
App
App
App
App
App
App OS OS OS
Operating System
Hypervisor
Hardware
Hardware
Traditional Stack
Virtualized Stack
Virtual Machines
Abstraction of a physical host machine
Hypervisor intercepts/emulates instructions; allows management

13. Virtualization Machines
Hardware
Virtual Machine Monitor (VMM) / Hypervisor VM
Windows
App
Linux
NetBSD
Xen
VMWare
UML
KVM
sHype
Hyper-V
Multiple virtual machines to run on a single physical machine
Performance: Para-virtualization (e.g., Xen) is very close to raw physical performance

14. Network Virtualization
Network virtualization is a method of combining the available resources in a network by splitting up the available bandwidth into channels, each of which is independent from the others, and each of which can be assigned (or reassigned) to a particular server or device in real time.

15. Characteristics of Cloud Computing01
Resource Pooling 02
Broad Network Access 03
On-Demand Self-Service 04
Rapid Elasticity 05
Measured Service

16. Economics of Cloud Users Pay by usage instead of provisioning for peak
Demand
Capacity
Time
Resources
Demand
Capacity
Time
Resources
Unused Resources

17. Need Programmable Networks
No practical way to experiment with new network protocols
GENI developing programmable networks that can support multiple isolated experimental networks
Focus here: how to run experiments in campus networks
Proposed approach: Obtain degree of switch flexibility:
High performance & low cost
Support broad range of research
Isolate experimental traffic from production traffic
Consistent with vendor’s need for closed platforms

18. OpenFlow Summary
Customization of the forwarding and treatment of classified traffic flows across the network.
Tight control over which packets are admitted into the network. .
OF specification is influencing
Availability of switches and routers that support OF
Design of packet-processing chips with advanced parsing and classification capabilities
Development of software-based switches

19. A Layered Architecture
SaaS
Portal
Custom KPIs
Urban Planning
Congestion pricing
3rd Party Apps …
BIaaS
Publish/Subscribe Overlay
Algorithmic Engines
Analytics Engines
APIs
PaaS
End-To-End, Multi Domain, Orchestration
Information-Centric Data Dissemination
Cloud Controllers
(SD) Network Controllers
Access/Things Controllers
SDI Resource
Management
SDI Manager
Topology Manager
Monitoring & Analytics
Software Defined
Infrastructure
Multi-Tier
Resources
Phys.

20. A Layered Architecture
SaaS
Portal
Custom KPIs
Urban Planning
Congestion pricing
3rd Party Apps …
BIaaS
Publish/Subscribe Overlay
Algorithmic Engines
Analytics Engines
APIs
PaaS
End-To-End, Multi Domain, Orchestration
Information-Centric Data Dissemination
Cloud Controllers
(SD) Network Controllers
Access/Things Controllers
SDI Resource
Management
SDI Manager
Topology Manager
Monitoring & Analytics
Software Defined
Infrastructure
Multi-Tier
Resources
Phys.

21. SAVI Multitier SDI Cloud
SAVI: Smart Applications on Virtual Infrastructure
NSERC Strategic Network
University & industry partnership
Application Platform on Multitier Cloud:
E2E orchestration of applications across multitier cloud
Core, Smart Edges, Access, vCPE, fog

22. SAVI Software-Defined Infrastructure
Integrated Management of Heterogeneous Resources
Computing, Networking, FPGAs, GPUs, SDR, Sensors, …
Virtual and Physical
Graph-based Topology Manager
Integrated measurement, monitoring & analytics

23. Janus Network Control Module
Port & Flavor info.
FlowStore
SDI Graph Topology Manager
Open APIs
Topology Info.
OpenFlow Events
Network Control & Mgmt Tasks
Graph Manager
Path Setup
Access Manager
ARP Handler
App Flow Manager
Network Controller Drivers
OpenFlow Controller 1
FlowStore Cache
OpenFlow Controller n
Nova/Neutron
Janus Plugins
No Broadcast
Routerless IP
Non-IP Traffic
Quality of Service
NFV Service Chaining
Security

24. Multi-Node Deployment
Four testbed-wide (i.e. centralized) services
Identity & Access Management
Web Portal
Image Registry
Orchestration Service
Currently 8 SAVI nodes are deployed in 7 Canadian Universities (1 Core + 7 Edges)
Each node essentially operates autonomously except a few testbed-wide services
Identity & Access Management (IAM)
Web Portal
Image Registry
Orchestration Service

25. SAVI Testbed Cross-Canada Testbed; L2 backbone
Carleton Edge
Cross-Canada Testbed; L2 backbone
Federated with US GENI, CloudLab & Chameleon
One SAVI node in Korea

26. Latest SAVI SDI Status
8 nodes In 7 Canadian Universities

27. SAVI SDI HW Resources
Cores
RAM (GB)
Storage (TB)
GPUs
3700+
8700+
440+ 7
FPGAs
SDRs
Wireless
IoT
60+ 15 23

28. SDN Advanced Features NFV Service Chaining Tapping Firewall/Blocking
Virtual Appliances and Physical Appliances
Tapping
Firewall/Blocking
Multicast
Distributed Virtual Router
Peering and Federation
L2 Wide Area Networking
NFV Service Chaining on Heterogeneous Cloud
SDN-based CPE

29. Service Chaining Gateway Web Server NFV (Snort DPI)
Versatile platform, rapid deployment
NFV (Snort DPI)

30. SDN Over Legacy Getting SDI Features on Legacy Platforms
Or Multiple Domains (Clouds)
Overlay Network Control
Using SAVI Janus SDI Manager
Web Server
(EC2)
Janus
Gateway
(EC2)
NFV (Snort DPI on SAVI)

31. End-To-End Orchestration
Global SDI Manager to Orchestrate across the testbed
Global SDI Manager
WAN
Manager
Local SDI Manager
Local SDI Manager
L2-Network
CANARIE/ORION
SAVI Core Node
Toronto
SAVI Edge Node
Waterloo

32. End-to-End Orchestration of L2
Physical OF Switch
Toronto Core
Controller
Agent
Waterloo Edge
Router VM
L2-Network
CANARIE/ORION
WAN
Manager
Global SDI Manager
Local SDI Manager

33. Federation with US Testbed/Clouds @L2
SDIXP

34. HyperExchange SDI eXchange Point
A versatile exchange point architecture that peers Infrastructure Providers (Autonomous SDI Domains) and their hosted VNs
Protocol agnostic
Based on SDI architecture
Inter-domain network control authorization
Enabling NFV at the edge of participating networks
US Testbed
Canada Testbed
VN1
VN2
Unified Logical VN
SDIXP
HyperExchange

35. SAVI CPE for IoT Demo SAVI Edge THINGS SAVI CPE UI Storage Computing …
Store-Forward
SAVI Edge
THINGS
SAVI CPE

36. SAVI vCPE Use Cases Gateway for Internet of Things devices
Web acceleration and service delivery point
NFVs such as proxy, firewall, IDS/IPS and VPN services
Smart home and office
Connected vehicles
Smart Transportation
Smart Cities (lighting, air quality, … , carbon footprint)

37. IoT Research Agenda IoT Virtualzation
Sensors, Actuators, Networks
SD Fabrics for City-Scale Infrastructure
Virtual Slices: Core + Smart Edge + vCPE + vIoT
Synergy with fiber-based broadband access
Synergy with wireless access: LTE, 5G, and more
IoT-scale data gathering and dissemination
Software-defined Information Centric Networking
Distributed storage, processing and aggregation
Security and Privacy
Low-latency and QoS where needed
Intelligence at Scale
Distributed analytics and deep learning

38. Conclusion
The SAVI multitier cloud based on SDI can provide flexibility, performance, scalability and cost effectiveness to support smart city applications
CVST application platform supports creation of smart transportation applications
Together SAVI & CVST provide a template for smart city application platforms

39. Thank You!