Re: HPDC Cloud Computing Panel Questions & My Answers

On Mon, Jun 9, 2008 at 4:42 PM, Greg Pfister <greg.pfister@gmail.com> wrote:

Q 1) What is and equally interestingly what isn't a cloud?

I would use ternary logic :)

 

Cloud:

...

 

Non-cloud:

...

 

Don't know:

...

 

 

 

Q 2) What is implication of Clouds for Enterprise Data Centers?

Lower cost, higher efficiency, downsizing.

 

in some cases total replacement of planned/existing DCs with cloud-based services (e.g. non-critical applications, research and development)

 

Q 3) If TeraGrid evolves to PetaCloud, how would it look?

Some major differences from Grid:
- inclusion of significant commercial infrastructure (databases, etc.)

- usually, a base assumption that the application is "transactional":
lots of separate requests, not one big manually-parallelized program.
This implicitly supports horizontal scaling in an automatic or near-
automatic manner. (Not required, but that's the usual target.)

- less need for heavily tuned performance in the client's code,
allowing use of scripting languages (Google Apps, others) in sandboxes
to code the app.

So a grid is unlikely to become a cloud, except of course for
marketing / funding buzzword purposes.

 

- compute/resource cycle units will have a fiscal equivalent

- grid will have tighter security model

- HPC guys will have to adopt new business model :) 

 

Q 4) What are research issues for Clouds?

Ease / simplicity of use. Scaling of all components. Performance
monitoring. Geographical failover. Possibly interoperability and
standards.

- specialization of clouds on certain type of works like image processing, CFD etc, for example recent development in GPU (NVidia and ATI) brings in very cheap hardware to perform parallel and matrix operations more efficiently (in some cases up to 10 times faster). Read more at http://www.nvidia.com/object/cuda_learn.html

 

Q 5) Can one (or who could) "trust" clouds?

We already do, inside corporate firewalls; clouds are already deployed
that way (customers invented clouds).

Outside the firewall – one must, to some degree, or the whole model
breaks. Legal implications for personal privacy may limit use outside
the firewall in some cases.

 

one should trust cloud, that operates in transparent mode

 

Q 6) Will Cloud interoperability be important; if so at what
interface(s) will it be provided?

It would certainly simplify life, but right how many providers are
happy with their walled gardens.

 

Cloud is a reflection of modern society, as long as there are more than two vendors, they will try to maintain it in their own way.

 

Q 7) What is the killer app for clouds?

The true "app" is lower cost of whole IT shops, hopefully coupled with
far better responsiveness. This alone could result in a major change
to how a wide swath of enterprise IT operates. All kinds of typical
commercial applications will be hosted – TWEaaS (The Whole Enchilada
as a Service).

However, narrowing the definition of "app" to what are traditionally
called applications, my personal favorite is cloud-level support for
virtual worlds.

All kinds of statistics point to a major uptake in the near future of
this technology. One recent one: Strategy Analytics estimates over the
next 10 years, 22% of all broadband users => over 1B people in one or
more virtual worlds. The infrastructure requirements for supporting
this are massive, heavily compute-intensive, and complex, requiring
the continuing simulation of the world.

This is a green-field area for IT shops that they will be more willing
to outsource to a cloud, as opposed to letting their current bread-and-
butter leave.

But most cloud vendors are as clueless as IT shops about this area, so
while it's low-hanging fruit, I don't know that they'll go after it.

 

This is a million dollar question :)

I think integration/adaption of existing infrastructure to clouds is area, where killer app might be found with very high probability.

 

 

 

Just my two cents.

 

 

cheers,

 

Khazret Sapenov,
Director of Research & Development
Enomaly Labs
US Phone: 212-461-4988 x5
Canada Phone: 416-848-6036 x5
E-mail: khaz@enomaly.com
Get Linked in> http://www.linkedin.com/in/sapenov

 

--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups "Cloud Computing" group.
To post to this group, send email to cloud-computing@googlegroups.com
To unsubscribe from this group, send email to cloud-computing-unsubscribe@googlegroups.com
For more options, visit this group at http://groups.google.ca/group/cloud-computing?hl=en
-~----------~----~----~----~------~----~------~--~---