« Return to Thread: Presentations & agenda
Re: Presentations & agenda
by Matt Mathis
::
Rate this Message:
There is another option to consider. If some portion of the GENI
infrastructure is placed at the RONs (Regional Optical Networks, aka GigaPoPs)
or at campus borders, it might be possible to "opt in" at very large
granularities by steering some portion of campus traffic through GENI. The key
here is to provide fast "restoration", e.g the capability to quickly reroute
traffic back on to production infrastructure. It also requires some fairly
strong incentive to encourage the users to opt in as a community.
There is good low hanging fruit - these days most campuses segregate dormitory
traffic from general university traffic, and subject traffic between the dorms
and external sites to separate bandwidth caps which are often much smaller
than the presented load. They do this to control the cost of providing
Internet service to the students, who have a huge capacity to share digital
content. The students would probably happily accept increased Internet
capacity in exchange for a slight reduction in robustness. The university
might save money by using a metered service for the students. For the cost of
providing transit service, GENI would have at its disposal a huge and diverse
source of authentic traffic.
The tricky parts are getting the "rerouting" to work well enough and to
accurately balance the desirability of the GENI service relative to the
parallel production service, such that the users continue to prefer to opt in
as a block, but they do not become so addicted that turning off GENI causes
problems. "Prime time down time" should probably be an ongoing operational
requirement.
Just a thought.
Thanks,
--MM--
-------------------------------------------
Matt Mathis http://www.psc.edu/~mathis
Work:412.268.3319 Home/Cell:412.654.7529
-------------------------------------------
Evil is defined by mortals who think they know
"The Truth" and use force to apply it to others.
On Wed, 3 Oct 2007, Henning Schulzrinne wrote:
> We have roughly 3 hours for the opt-in WG meeting in Minneapolis.
> After discussions with Aaron, we will be spending a good chunk of our
> time trying to scope the problem. One way to divide the topic into
> more manageable chunks is to think of a division into three problems:
>
> (1) "ISP": If GENI acts like an ISP, how do users gain access to
> that, both on traditional Ethernet-style (campus) networks and on
> wireless networks (802.11 and cellular)? What incentives are there?
>
> (2) "Generalized P2P": If end user resources become part of the
> experimental infrastructure, how can users restrict the use of such
> resources and how can experimenters access those resources?
>
> (3) Network services: How do users get access to network services?
> What are the motivations and possible incentive/payment mechanisms?
>
> All three problems have both a technical component, such as protocols
> and issues of trust, an economic component (who pays and in what
> currency) and a policy/legal component, such as the usual issues that
> worry institutional review boards (IRBs), including privacy and
> informed consent.
>
> To help structure the discussion, I'd like to invite members of the
> WG to present brief talks on WG-related topics, not necessarily
> following the structure above. Please let me know as soon as possible
> what you would want to talk about, so that I can put together an agenda.
>
> Henning
>
> _______________________________________________
> opt-in-wg mailing list
> opt-in-wg@...
> http://lists.geni.net/mailman/listinfo/opt-in-wg
>
_______________________________________________
opt-in-wg mailing list
opt-in-wg@...
http://lists.geni.net/mailman/listinfo/opt-in-wg
infrastructure is placed at the RONs (Regional Optical Networks, aka GigaPoPs)
or at campus borders, it might be possible to "opt in" at very large
granularities by steering some portion of campus traffic through GENI. The key
here is to provide fast "restoration", e.g the capability to quickly reroute
traffic back on to production infrastructure. It also requires some fairly
strong incentive to encourage the users to opt in as a community.
There is good low hanging fruit - these days most campuses segregate dormitory
traffic from general university traffic, and subject traffic between the dorms
and external sites to separate bandwidth caps which are often much smaller
than the presented load. They do this to control the cost of providing
Internet service to the students, who have a huge capacity to share digital
content. The students would probably happily accept increased Internet
capacity in exchange for a slight reduction in robustness. The university
might save money by using a metered service for the students. For the cost of
providing transit service, GENI would have at its disposal a huge and diverse
source of authentic traffic.
The tricky parts are getting the "rerouting" to work well enough and to
accurately balance the desirability of the GENI service relative to the
parallel production service, such that the users continue to prefer to opt in
as a block, but they do not become so addicted that turning off GENI causes
problems. "Prime time down time" should probably be an ongoing operational
requirement.
Just a thought.
Thanks,
--MM--
-------------------------------------------
Matt Mathis http://www.psc.edu/~mathis
Work:412.268.3319 Home/Cell:412.654.7529
-------------------------------------------
Evil is defined by mortals who think they know
"The Truth" and use force to apply it to others.
On Wed, 3 Oct 2007, Henning Schulzrinne wrote:
> We have roughly 3 hours for the opt-in WG meeting in Minneapolis.
> After discussions with Aaron, we will be spending a good chunk of our
> time trying to scope the problem. One way to divide the topic into
> more manageable chunks is to think of a division into three problems:
>
> (1) "ISP": If GENI acts like an ISP, how do users gain access to
> that, both on traditional Ethernet-style (campus) networks and on
> wireless networks (802.11 and cellular)? What incentives are there?
>
> (2) "Generalized P2P": If end user resources become part of the
> experimental infrastructure, how can users restrict the use of such
> resources and how can experimenters access those resources?
>
> (3) Network services: How do users get access to network services?
> What are the motivations and possible incentive/payment mechanisms?
>
> All three problems have both a technical component, such as protocols
> and issues of trust, an economic component (who pays and in what
> currency) and a policy/legal component, such as the usual issues that
> worry institutional review boards (IRBs), including privacy and
> informed consent.
>
> To help structure the discussion, I'd like to invite members of the
> WG to present brief talks on WG-related topics, not necessarily
> following the structure above. Please let me know as soon as possible
> what you would want to talk about, so that I can put together an agenda.
>
> Henning
>
> _______________________________________________
> opt-in-wg mailing list
> opt-in-wg@...
> http://lists.geni.net/mailman/listinfo/opt-in-wg
>
_______________________________________________
opt-in-wg mailing list
opt-in-wg@...
http://lists.geni.net/mailman/listinfo/opt-in-wg
« Return to Thread: Presentations & agenda
| Free embeddable forum powered by Nabble | Forum Help |
