[ovs-discuss] My M.Sc. thesis report and our IEEE NetSoft 2015 Conference paper

[Prashanth Varma Manthena]

Dear Mr. Pfaff,

First of all, thank you for your quick response and wishes.

Moreover, I successfully defended my thesis and graduated with a good final
grade from the university (TU Delft, The Netherlands) in February this
year.

All I have to say is that the Open vSwitch project is driving the
revolution and innovation in network architectures and services.
Furthermore, It is making SDN and NFV based research much more accessible,
open, and collaborative.

I personally was able to test, implement, and validate SDN
architectures successfully against real-world networks using OVS switches.
OVS physical switches with hardware (ASIC) support for IP/MPLS operations
were able to perform as good as any vendor-specific (legacy) switches and
routers. Here are my findings (refer second paragraph in pg. 52 of my
thesis):

During the performance evaluation experiments on the PoC physical network
testbed, it was found that Testbed Setup B involves slightly higher
round-trip delay/time (RTD/RTT) compared to that of Testbed Setup A (by
around 100 milliseconds). The main reason behind this is that the two open
switches, 48 port 1 GbE Pica8 P-3290 open switches, in Testbed Setup B’s
network core support MPLS switching operations (push, pop, and swap) in
their CPU instead of their hardware ASICs. Thus, the involved open switches
in the proposed evolutionary approach must support MPLS switching
operations in their hardware ASICs for much faster switching, avoidance of
network bottlenecks, and to prevent unnecessary load on the switch CPUs.

I have couple of direct advises and suggestions to the Open vSwitch project:

1) Compare the performance of different implementations (software,
physical, and hardware (ASIC) support) of OVS switches . Such information
would be very useful and inspirational for service providers and network
operators, who are planing to implement SDN and NFV  based architectures
for their production and real-world networks.

2) Push towards open-standards based network configuration protocols with
fully-programmable interfaces to realize overall virtualization of
networks. From my experiences, most of the vendors implementing OVS in
their switches still only support vendor-specific CLI and web-based network
configuration, this will change with much more support for network
configuration protocols like NETCONF and OF-CONFIG along wth OVSDB from the
Open vSwitch project.


Let me know, if you need any additional information from my side.


With regards,
*M.P.V. Manthena (Prashanth)*

M.Sc. in Electrical Engineering
Delft University of Technology (TU Delft), The Netherlands

LinkedIn Profile : http://nl.linkedin.com/in/prashanthvarma/
Personal Website: http://prashanthvarma.com/

On Mon, Apr 6, 2015 at 6:58 PM, Ben Pfaff <blp at nicira.com> wrote:

> On Mon, Apr 06, 2015 at 04:56:58PM +0200, Prashanth Varma Manthena wrote:
> > I just want to update you all with the information regarding my M.Sc.
> > thesis report and our IEEE NetSoft 2015 Conference paper.
>
> Congratulations on your paper being accepted and (presumably) your
> graduation.
>
> I skimmed through your conference paper.  Do you have any direct
> advice or suggestions for the Open vSwitch project?
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://openvswitch.org/pipermail/ovs-discuss/attachments/20150406/162623cc/attachment-0002.html>