[vnrg] IETF Virtual network

[Vishwas Manral <vishwas.ietf at gmail.com>]

Hi Sunay/ Robert,

I looked at your presentation at the IETF. You seem to talk about
cloud but seem to be talking more specfically about IaaS.

1. Most current cloud providers (Amazon, Microsoft) charge based on
the number and duration of instances in the servers. Google is the
only one for its AppEngine charges on basis of CPU.

2. Amazon has the Route53 - DNS service, for which they charge.

3. I am not sure what you mean by things like MAC-in-MAC becomes
complicated? It is an existing functionality. TRILL (of which I am one
of the authors) is something that seems to be in development by most
network device providers.

4. Cloud customers most want security and protection of data. (I can
exchange the link where I found this data) They then want reliablity,
performance and cost. Besides that there are issues with regulatory
concerns.

5. We can use VLAN, or L2 ACL's for partitioning however there can be
issues of spoofing in some cases.

6. I do not see a reason why we will not allow an operator to Layer-2
virtual network over the substrate layer.

7. I think though OpenFlow is provided, I know most vendors are
working on/ already provided providing open API's over which anything
can run not just OpenFlow.

8. You have not talked about QoS at all in the presentation.

9. Though I agree SNMP can be a way to program varied devices, I think
NetConf or something similar based on XML will be more helpful.

Thanks,
Vishwas