CSO Architecture, Terminology, Separable Roles...
Greg Bernstein
Sorry to take so long to get around to reviewing this document. There
are a number of nice concepts and terminology here that get one
thinking (always dangerous :-)).
First I like the notion of the split between "Application", "IT
resource" and "Network Resource". Here is my attempt at defining these
along with your definitions:
"Network Application": The service being delivered to an "end user"
across some type of network. We exclude strictly local applications
since they do not use "network" resources. A "Network application" by
its nature consumes both "Network resources" and "IT resources". It
seems we also want to include necessary support tasks such as backup,
data replication since these can be resource consumers (but may also
be more flexible in their delivery).
"IT Resources": (your definition) in the context of this document, the
Information Technology (IT) resources refer to computing and storage
resources (disk capacity, CPU, etc) typically residing on Data Centers
(DC) spread along the network. I like this definition since
"application resource" is a bit too general (i.e., includes network)
and "compute resource" is a bit too narrow (doesn't sound like it
includes storage). Would we include the internal (data center) network
connecting these resources together?
"Network Resources": Raw network resources and end to end services
with QoS guarantees. Some of us are concerned with rather large pipes,
i.e., wavelength granularity (10-40Gbps +) others smaller with less.
In addition we have a number of networking layers that we may involve,
i.e., we consider layers below IP!
Separable roles that may be involved in "Network application" delivery
(many entities combine some or all these roles):
"Application Provider": (rough) the folks ultimately responsible for
the application. E.g., A game company, video content provider, etc...
"IT Resource Providers" (Data Centers): The folks that provide "IT
resources", e.g., cloud providers, 3rd party data centers, etc...
"Network Resource Providers": The ISPs, carriers, etc... that provide
the networks that carry the application traffic.
Other main roles? What would each role knows (or could know) and what
might they be willing to share for optimization/resiliency purposes?
Cheers
Greg B.
LUIS MIGUEL CONTRERAS MURILLO
Thanks for your valuable comments. I've inserted my own comments in line (they only reflect my personal view).
Best regards,
Luis
-----Mensaje original-----
De: cross-stratum...@googlegroups.com [mailto:cross-stratum...@googlegroups.com] En nombre de Greg Bernstein
Enviado el: jueves, 01 de diciembre de 2011 19:44
Para: Cross Stratum Optimization
Asunto: {Cross Stratum Optimization} CSO Architecture, Terminology, Separable Roles...
Hi Luis and co-authors of http://www.ietf.org/id/draft-contreras-cso-functional-architecture-00.txt.
Sorry to take so long to get around to reviewing this document. There are a number of nice concepts and terminology here that get one thinking (always dangerous :-)).
First I like the notion of the split between "Application", "IT resource" and "Network Resource". Here is my attempt at defining these along with your definitions:
"Network Application": The service being delivered to an "end user"
across some type of network. We exclude strictly local applications since they do not use "network" resources. A "Network application" by its nature consumes both "Network resources" and "IT resources". It seems we also want to include necessary support tasks such as backup, data replication since these can be resource consumers (but may also be more flexible in their delivery).
"IT Resources": (your definition) in the context of this document, the Information Technology (IT) resources refer to computing and storage resources (disk capacity, CPU, etc) typically residing on Data Centers
(DC) spread along the network. I like this definition since "application resource" is a bit too general (i.e., includes network) and "compute resource" is a bit too narrow (doesn't sound like it includes storage). Would we include the internal (data center) network connecting these resources together?
"Network Resources": Raw network resources and end to end services with QoS guarantees. Some of us are concerned with rather large pipes, i.e., wavelength granularity (10-40Gbps +) others smaller with less.
In addition we have a number of networking layers that we may involve, i.e., we consider layers below IP!
Separable roles that may be involved in "Network application" delivery (many entities combine some or all these roles):
"Application Provider": (rough) the folks ultimately responsible for the application. E.g., A game company, video content provider, etc...
"IT Resource Providers" (Data Centers): The folks that provide "IT resources", e.g., cloud providers, 3rd party data centers, etc...
"Network Resource Providers": The ISPs, carriers, etc... that provide the networks that carry the application traffic.
Other main roles? What would each role knows (or could know) and what might they be willing to share for optimization/resiliency purposes?
Regarding to what extent the information could be shared for optimization or resiliency purposes, I guess that both network operators and DC providers would not be willing to transparently offer such kind of information. Probably, the optimization process would be based on trusted transactions between both, where both could offer information about the resources at the borders, but hiding the internal information,in such a way that the internal optimization (to DC and to the network) is left to each of them.
Cheers
Greg B.
Este mensaje se dirige exclusivamente a su destinatario. Puede consultar nuestra política de envío y recepción de correo electrónico en el enlace situado más abajo.
This message is intended exclusively for its addressee. We only send and receive email on the basis of the terms set out at.
http://www.tid.es/ES/PAGINAS/disclaimer.aspx
Greg Bernstein
CSO Wiki at http://cccso.net/mw/index.php/Terminology and I'll update
with your comments.
Also (all), I did put up an initial summary of some of the background
references cited on this list and a few that I like at
http://cccso.net/mw/index.php/Background. Complete references for
items cited below can be found there.
See below for more discussion.
Cheers
Greg
On Dec 5, 4:01 am, LUIS MIGUEL CONTRERAS MURILLO <l...@tid.es> wrote:
> Hi Greg,
>
> Thanks for your valuable comments. I've inserted my own comments in line (they only reflect my personal view).
>
> Best regards,
>
> Luis
>
-- Snip --
>
> [Luis>>] Agree. Probably it is more proper to talk about "NetworkED application"; that is, an application that needs or uses network interconnection to reach its goals. Furthermore, the "end user" could be understood as either an operator subscriber (or group of subscribers), a data center controller, a network entity (e.g., and ALTO server), etc.
--> Yes. I don't mean to use the term "end user" in a restrictive way.
When I was getting back up to speed on the literature the paper
[Ratnasamy02] (see CSO Wiki Background page) talks about using
approximate network information to help in both the "server selection
problem" and "overlay construction". Where by overlay construction
they mean the interconnection of "IT resources". Hence I think of this
type of "overlay construction" as included in a "networkED
application" and particularly useful in applications involving mutiple
data centers (3rd part, corporate, etc...).
>
> "IT Resources": (your definition) in the context of this document, the Information Technology (IT) resources refer to computing and storage resources (disk capacity, CPU, etc) typically residing on Data Centers
> (DC) spread along the network. I like this definition since "application resource" is a bit too general (i.e., includes network) and "compute resource" is a bit too narrow (doesn't sound like it includes storage). Would we include the internal (data center) network connecting these resources together?
>
> [Luis>>] This is a good question. I think that the border between "IT resources" and "Network resources" is defined by the controller which manages the corresponding resources. Due that the internal data center network (switches) are typically controlled by the DC controller, these resources are part of the "IT resources" set. The key point is who controls the device providing connectivity to the network (the WAN router) to determine where the border is. In any case, an operator could go further and also control the switches of the data center, reducing the scope of the IT resources to the more computing related devices (storage, CPU, etc). I think this border can be flexible.
>
--> Good point. Agree.
-- snip --
>
> Separable roles that may be involved in "Network application" delivery (many entities combine some or all these roles):
>
> "Application Provider": (rough) the folks ultimately responsible for the application. E.g., A game company, video content provider, etc...
>
> "IT Resource Providers" (Data Centers): The folks that provide "IT resources", e.g., cloud providers, 3rd party data centers, etc...
>
> "Network Resource Providers": The ISPs, carriers, etc... that provide the networks that carry the application traffic.
>
> Other main roles? What would each role knows (or could know) and what might they be willing to share for optimization/resiliency purposes?
>
> [Luis>>] Some other new roles could emerge if we consider issues like resource virtualization, either for IT or for the network worlds. For instance, in FP7 GEYSERS project different roles have been defined (you can see chapter 2 of GEYSERS public deliverable 1.1, downloadable athttp://www.geysers.eu/images/stories/deliverables/geysers-deliverable...).
--> Thanks for the reference. This is good information and illustrates
more general optimization than just the server selection problem. The
GEYSERS website also provides some motivation/background on the
importance/use of optical networks.
> Regarding to what extent the information could be shared for optimization or resiliency purposes, I guess that both network operators and DC providers would not be willing to transparently offer such kind of information. Probably, the optimization process would be based on trusted transactions between both, where both could offer information about the resources at the borders, but hiding the internal information,in such a way that the internal optimization (to DC and to the network) is left to each of them.
--> Hmm, in GMPLS and some of the UNIs (or UNI proposals) we provide
for specification of protection levels of underlying links
(unprotected, linear, ring, etc...) or for individual connections
(LSPs). It seems we should be able to abstract this in a suitable form
for sharing with "somewhat trusted" partners.
>
-- snip --
LUIS MIGUEL CONTRERAS MURILLO
Thanks for your inputs. One last comment:
> Regarding to what extent the information could be shared for optimization or resiliency purposes, I guess that both network operators and DC providers would not be willing to transparently offer such kind of information. Probably, the optimization process would be based on trusted transactions between both, where both could offer information about the resources at the borders, but hiding the internal information,in such a way that the internal optimization (to DC and to the network) is left to each of them.
--> Hmm, in GMPLS and some of the UNIs (or UNI proposals) we provide
for specification of protection levels of underlying links (unprotected, linear, ring, etc...) or for individual connections (LSPs). It seems we should be able to abstract this in a suitable form for sharing with "somewhat trusted" partners.
Best regards,
Luis