1. معرفی شبکه های مبتنی بر نرم افزار Software Defined Networks
دانشگاه علم و صنعت ایران
دانشکده مهندسی کامپیوتر
دکتراحمد اکبری – دکتر محمود فتحی – دکتر حسین غفاریان
مهندس شتابی
بهمن 1393

3. دانشگاه علم و صنعت ایران دانشکده مهندسی کامپیوتر
تاسیس: 1368
تعداد اعضای هیات علمی
2 استاد تمام
10 دانشیار
8 استادیار
4 کارشناس آموزش
تعداد فضاهای تحقیقاتی با موضوعات شبکه: 9 آزمایشگاه تحقیقاتی
+ مرکز تحقیقات فناوری اطلاعات
دانشکده مهندسی کامپیوتر دانشگاه علم و صنعت ایران

4. زمینه های همکاریهای گذشته و فعلی با توانیر
فاز نخست بازطراحی WAN صنعت برق ( )
ارائه طرح کلان شبکه WAN داده صنعت برق با توجه به نیازهای فعلی (در حال اجرا)
عضویت در کمیته داده صنعت برق به عنوان مشاور دانشگاهی کمیته
دانشکده مهندسی کامپیوتر دانشگاه علم و صنعت ایران

5. فناوریهای نوین حوزه شبکه های کامپیوتری
Cloud computing
Software Defined Network
دانشکده مهندسی کامپیوتر دانشگاه علم و صنعت ایران

6. رایانش ابری Cloud computing
دانشکده مهندسی کامپیوتر دانشگاه علم و صنعت ایران

7. سرویس های رایانش ابری
دانشکده مهندسی کامپیوتر دانشگاه علم و صنعت ایران

8. پیش بینی سیسکو از روند رشد مجموع ترافیک مراکز داده در سالهای 2013 تا 2018
دانشکده مهندسی کامپیوتر دانشگاه علم و صنعت ایران

9. رشد ابزارهای متصل به اینترنت
Internet of things ( Cloud Networks  Fog Networks)
دانشکده مهندسی کامپیوتر دانشگاه علم و صنعت ایران

10. اهداف شبکه های مبتنی بر نرم افزار SDN
کاهش هزینه های عملیاتی شبکه
انعطاف پذیری در حوزه مدیریت شبکه
کاهش زمان بروز رسانی در مجموعه ابزارهای شبکه
مدیریت سراسری و یک پارچه شبکه
کاهش هزینه های سخت افزارهای ارتباطی شبکه
دانشکده مهندسی کامپیوتر دانشگاه علم و صنعت ایران

11. معماری مسیریابها و سوئیچهای فعلی
Specialized Packet Forwarding Hardware
Operating
System
Feature
دانشکده مهندسی کامپیوتر دانشگاه علم و صنعت ایران

12. توزیع شدگی روندهای کنترلی در معماری فعلی ابزارهای شبکه
دانشکده مهندسی کامپیوتر دانشگاه علم و صنعت ایران

13. چهره جدید شبکه
دانشکده مهندسی کامپیوتر دانشگاه علم و صنعت ایران

14. Specialized Packet Forwarding Hardware
معماری شبکه های کامپیوتری
Closed Platform
Open Platform
3rd party App
3rd party App
Feature
Feature
Public APIs
Proprietary interface
OS / Controller
Kernel OS +
Specialized Packet Forwarding Hardware
Standard hardware
سخت افزار اختصاصی
رابط اختصاصی
سرعت نواوری کند
سخت افزار استاندارد
استانداردهای باز – سیستم پویا
سرعت نواوری سریع
معماری شبکه های SDN
معماری شبکه های سنتی 14

15. Control / Data plane separation
App
App
App
Unified Data and Control
Control + Data Separation
Add feature here?!?!
Controller
versus
معماری شبکه های SDN
Control
معماری شبکه های سنتی
Data
روش سنتی، سوئیچ / روتر مسئول تصمیم گیری سطوح داده و کنترل هستند
سطوح کنترل و دیتا از هم جدا شده وتصمیم گیری و مسیر یابی توسط کنترلر انجام می شود و سوئیچ کار ارسال دیتا را دارد. 15

16. The network is changing
Feature
Network OS
Feature OS
Feature
Custom Hardware OS
Feature
Custom Hardware OS
Feature
Custom Hardware OS
Custom Hardware
Feature OS
Custom Hardware

17. Software Defined Network (SDN)
2. At least one Network OS probably many. Open- and closed-source
3. Consistent, up‐to‐date global network view
Feature
Feature
Network OS
1. Open interface to packet forwarding
Packet
Forwarding
Packet
Forwarding
Packet
Forwarding
Packet
Forwarding
Packet
Forwarding

18. Network OS
Network OS Examples

19. Network OS
Network OS Examples

20. Software Defined Network (SDN)

21. SDN - Layered Abstraction
Application
Layer
Deliver open programmable interfaces to automate orchestration of network services
Control
Layer
Separate control and data plane; abstract control plane of many devices to one
SDN Architecture
Define SDNs as requiring the following three elements:
A method for modifying packet forwarding rules and/or applying policy to packets
A method for doing element 1 across multiple devices, i.e. applying packet forwarding rules and polices across multiple devices (not just one device) in a dynamic and coordinated fashion
The ability to perform elements 1 and 2 in a programmable fashion
Multiple layers of API possible
Access to services that ‘compile’ the complexity of the actual network into usable components; e.g. topology; these are the ‘libraries’ for the programmability
Infrastructure Layer
Open standard-based programmatic access to infrastructure

22. SDN - Layered Abstraction
Application
Layer
Network Applications
SDN Applications
Business Applications
(e.g., OpenStack, CloudStack)
Cloud Orchestration
Deliver open programmable interfaces to automate orchestration of network services
Deliver open programmable interfaces to automate orchestration of network services
Deliver open programmable interfaces to automate orchestration of network services
SDN Controller
Programmable Open APIs
Control
Layer
Separate control and data plane; abstract control plane of many devices to one
Separate control and data plane; abstract control plane of many devices to one
SDN Architecture
Define SDNs as requiring the following three elements:
A method for modifying packet forwarding rules and/or applying policy to packets
A method for doing element 1 across multiple devices, i.e. applying packet forwarding rules and polices across multiple devices (not just one device) in a dynamic and coordinated fashion
The ability to perform elements 1 and 2 in a programmable fashion
Multiple layers of API possible
Access to services that ‘compile’ the complexity of the actual network into usable components; e.g. topology; these are the ‘libraries’ for the programmability
Infrastructure Layer
Open standard-based programmatic access to infrastructure
Open standard-based programmatic access to infrastructure
Network Device
Control & Data Plane Programmable
Interface (e.g., OpenFlow)

23. SDN - Layered Abstraction

24. OpenFlow Basics

25. OpenFlow Basics

26. OpenFlow Basics

27. OpenFlow Basics
OpenFlow Flow Table Entry

28. OpenFlow Basics
Flow Table Examples

29. OpenFlow Basics
Flow Table Examples

30. OpenFlow Basics
How a packet is processed and forwarded in an OpenFlow switch

31. OpenFlow Basics

32. OpenFlow Basics

33. OpenFlow Basics

34. OpenFlow Basics
OpenFlow-only switches: support only OpenFlow operation

35. OpenFlow Basics
Comparison of OpenFlow Specifications

36. OpenFlow Basics
Industry Standards and Forums

37. OpenFlow Basics
Example of OpenFlow-Compliant Switches

38. SDN Applications Scenarios

39. SDN Applications Scenarios
Scenario 1: Load balancing for better network usage
Use cases - Google
Before using SDN/OpenFlow, network utilization rate ~ 30%
After adopting SDN/OpenFlow, network utilization rate above 90%
Reason: Without global network information, many flows may be unevenly routed to certain segments of the network causing network congestion
SDN based network can route flows based on the global NIB to intelligently distribute the flows evenly on all available segments – increased network utilization
Google applied SDN/OpenFlow in their WAN and data center network

40. SDN Applications Scenarios
Current network taking long time for propagate events to every network nodes
SDN/OpenFlow adjusting network updates at much faster (milliseconds) speed – maximizing network usability and throughputs
Scenario 2: Fast network adaption and traffic allocation
Use case: Different time using different links for better network usage and OPEX (Policy based routing)
At day time work hours, route traffic to shortest, low latency path (SLIP), and route the traffic to longer (cheaper) alternative path (LAP) during the night and off working hours

41. SDN Applications Scenarios
Use case A: Energy saving network (Green network)
Distribute network traffic evenly for maximum throughputs/low latency during day time working hours,
Distribute network traffic to certain segments and shut-down others for energy saving during late night off peak hours
Dynamically activate network segments when network traffic picking up
Use case B: Application Driven network
Adjusting on-demand to match application needs
With less number of applications running, the network traffic can be concentrated to few links and intelligently re-distribute the network traffic when needed for better QoS
Dynamically configure network to fit applications with application dependent flow based routing policy for optimal performance

42. SDN Applications Scenarios
Scenario 4: Unified network control and network flattening
SDN/OpenFlow routing vector dimension increased from 15 (OF 1.0) to 36 (OF1.3) and extensible, covering routing parameters on multiple layers (L2/L3/L4/etc.)
Data networking could be flattened – one path may achieve routing on multiple layers – reducing network complexity/COPEX, increasing network efficiency, etc.

43. SDN Applications Scenarios
Scenario 5: Open and programmable network, optimized for applications
SDN provides southbound API for data forwarding and northbound API do network applications – allowing network be changed by application programs as software to optimally support applications
Use case A: Programmable network resources
Network resources can be programmed to allocate/release , e.g. on-demand bandwidth, etc. allocated and adjusted
Applications can programmability optimize their resource utilization
Use case B: Automated network management
Programmable network can reduce the complexity and cost of network deployment and management, increase the speed of service deployment, reduce human involvement and error, make network management more automatic, etc.

44. SDN Applications Scenarios
End-to-End QoS
SDN is based on logically centralized network control and can achieve globally optimized control – making end-to-end QoS possible
Fast network adjustment and recovery
SDN can timely provide alternative optimal path, avoid network trouble spots with fast network recovery at the speed of milliseconds
Multi-network federation
SDN/OpenFlow based core network can normalize and synchronize network signaling in multi-network federation (e.g. tagging normalization, etc.）

45. Software Defined Networking course https://www.coursera.org
Dr. Nick Feamster
Georgia Institute of Technology
Starts in 3 months
Eligible for Verified Certificate
8 weeks of study
7-10 hours/week
English
دانشکده مهندسی کامپیوتر دانشگاه علم و صنعت ایران

46. Software Defined Networking https://www.coursera.org/course/sdn
Course Syllabus
This course will cover 8 modules (one per week).
Module 1: History and evolution of SDN
Module 2: Control and data plane separation
Module 3: Control Plane
Module 4: Network Virtualization
Module 5: Data Plane
Module 6: Programming SDNs
Modules 7: Verification and Debugging
Module 8: Use Cases and Looking Forward
دانشکده مهندسی کامپیوتر دانشگاه علم و صنعت ایران

47. با تشکر از حسن توجه شما
دانشکده مهندسی کامپیوتر دانشگاه علم و صنعت ایران