nanogf . nanogf at
Fri Sep 24 21:10:27 UTC 2010

I plan to distribute OpenFlow switches, which is why I would like to further dis
cuss about this technology with NANOGers.
Firstly, to get a better overview of OpenFlow,  I would greatly appreciate to in
vite you to a further reading of the latest presentation on the subject from NEC
 entitled "Future Internet Research Using OpenFlow" [1].
I) Definition
a) Commercial
I'd note that Flow-based networking is not a new concept and is currently suppor
ted by Anagran [2], whose founder is none other than Larry Roberts [3].
b) Dynamic Switch Control Protocols
By the way,  I would greatly appreciate to invite you to a further reading of th
e thesis entitled "Implementation and Evaluation of a Network Element Control Pr
otocol" [4] where different Dynamic Switch Control Protocols (Strengths, and GSM
P OpenFlow) are compared (page 17-23).
-Note: the source code is available at the web page titled "Work ETNA Package 4
II) Business Model
1) Capex
The advantage of OpenFlow is to separate the Data Plane from the Control Plane,
thus simplifying the switches (see page 15 of "Future Internet Research Using Op
enFlow "[0]) which results in a reduction of Capex (see page 23 of the presentat
ion entitled "OpenFlow / Software Defined Networks" [6]).
>From a cost-perspective you can deploy a cheaper networking gear basedon "dummy
 OpenFlow-only" switches, provided a controller applicationthat takes care of th
e network intelligence (e.g., installing flows,VM directory, etc.).
To illustrate this argument, I would greatly appreciate to invite you to a furth
er reading of the presentation entitled "OpenFlow / Software Defined Networks" [
6] from Clean Slate project at Stanford and the one entitled "Data Center Networ
ks Are in My Way" [7] from Cloud Provider, Amazon.
2) Opex
Moreover, intelligence is centralized through the OpenFlow controller therefore
easier, resulting in a reduction of Opex (see page 26 of the presentation entitl
ed "OpenFlow / Software Defined Networks" [6]).
>From a functional perspective you have the potential gains ofre-thinking the pa
cket forwarding departing from spanning trees orVLAN segmentations (cf. SIGCOMM0
9 Portland paper).
To illustrate this second argument,  I would greatly appreciate to invite you to
 a further reading of the presentation entitled "An Experimenter's Guide to Open
Flow" [8] of Deutsche Telekom R&D Labs and the one entitled "OpenFlow : Operatio
nal Experiences" [9] from the GRNOC at the University of Illinois (who will also
 give an OpenFlow track at NANOG50 [10])
-Note: as an example of research application, I would greatly appreciate to invi
te you to a further reading of the presentation entitled "Towards a Flow-level N
etwork Security System "[11].
III) Open Source
For those wishing to experiment with this technology, I would advise the followi
ng setup:
-HW :
Pronto 3290 [12]
-Firmware :
Pica8 XorPlus 1.1 that includes the L2/L3 management for VLAN, LACP, STP/RSTP, L
LDP, OSPF, RIP, static route, PIM-SM, VRRP, IGMP, IGMP Snooping, IPv6, Radius/Ta
cacs+ as well as OpenFlow 1.0 (available at later Oct, 2010) [13]
-OpenFlow-Controller :
Beacon: Java OpenFlow Controller [14]
Have a good week-end :) !
I look forward to your answer,
Best Regards,

[1] [1]
[2] [2]
[3] [3]
[4] [4]
[5] [5]
[6] [6]
[7] [7]
[8] [8]
[9] [9]
[10] [10]
[11] [11]
[12] [12]
[13] [13]
[14] [14]

Guillaume FORTAINE
Tel : +33(0)631092519
Mail : gfortaine at


   Get your own *free* email address like this one from



More information about the NANOG mailing list