Hi,

As the author of rfc5706, and a long-time participant in the OPS area, let
me make some comments and provide some advice.

First, 
OAM and "Operations and Management" are not the same thing.
There is an RFC that discusses OAM terminology, and I recommend reading it
to understand the distinction (which still is NOT unambiguous).

Generally, 1) OAM happens inline as part of the protocol, 2) operations
includes consideration of the environment in which the protocol will
operate, and 3) management tends to be a combination of layer 7 protocols
that provide monitoring/management capabilities combined with
protocol-layer-specific information, such as MIB objects or ipfix
information elements or syslog structured data, etc. IETF management
protocols typically keep a clear distinction between the protocols used to
carry management information and the data models used to model the managed
functionality. 

Second, 
as the WG starts to consider how to address O&M requirements and/or OAM
requirements, I recommend starting with an understanding of the difference
between information models and data models, as described in RFC3444 (IIRC).

OPSAWG is doing a lot of work at making sure the information used by one
protocol can be correlated and/or translated between data models for use
with other management protocols.

I recommend developing a management-protocol-independent information
model, plus at least one management-protocol-specific data model (which
helps serve as a proof that the information model is viable).


Third,
The IETF currently has four main standards in the "management protocol
toolkit": SNMP (monitoring and tweaking), IPFIX (flow monitoring), Netconf
(configuration), and Syslog (logging). These are widely supported by
operators, and standard data models should be available so operators can
use these IETF management protocols to manage IETF protocols when possible.

Fourth,
RFC5706 discusses things to consider when developing a management
information model (and subsequent data models).
For example, for ALTO, an operator might want to be able to monitor what
sources are being used to gather physical network information, what
transform algorithms are being applied to the source data to produce ALTO
application-level information, and to monitor who is getting that
processed information from the ALTO server.
You might want to consider whether protocol errors are being detected
between the server and the data sources, or between the server and
specific consumers. This might be important to understand whether certain
sources or consumers are putting an unnecessary strain on the server, so a
server might choose to stop communicating with that error-generating
connection (and whether that decision should be standardized). That
infomation also might help detect why certain consumers might choose
non-optimal optimizations.

RFC5706 also discusses operational requirements. If BGP will be used to
gather source data, then BGP must be supported in the environment, and
that has operational considerations. Will all ALTO servers operate in BGP
environments? If an operator decides BP is not important to the physical
network they run, or choose not tp provide BGP data for security reasosn,
how will the ALTO environment be affected, and will this affect
interoperability across ALTO servers?

And so on ...

--
David Harrington
Internet Engineering Task Force (IETF)
[email protected]
+1-603-828-1401





On 5/11/12 12:30 AM, "Martin Stiemerling" <[email protected]>
wrote:

>Hi there,
>
>An important note on the ALTO protocol as part of the WGLC:
>
>The ALTO requirements have been discussed in the 2012-05-10 IESG
>telechat and one point raised which is actually of importance for the
>ALTO protocol is this:
>
>There is/was a DISCUSS on the ALTO requirements about operations and
>management (OAM) of the ALTO protocol.
>RFC 5706 (http://tools.ietf.org/html/rfc5706) gives an overview of
>points to be checked for new protocols.
>
>However, we have not considered RFC 5706 (at least to my knowledge) in
>the ALTO protocol that discusses operations and management of the ALTO
>protocol.
>
>The outcome of the IESG discussion is that the ALTO protocol must cover
>the aspect of OAM and discuss what is potentially needed and what is
>potentially not needed.
>
>For instance:
>- the support of logging at the server for failure detection is a
>potentially a good idea, though it might become touchy in the p2p use
>case.
>- discussion of what clients are supposed to do when a server fails for
>various reasons. This is probably already covered in the current ALTO
>protocol draft.
>
>Regards,
>
>   Martin
>
>-- 
>IETF Transport Area Director
>
>[email protected]
>
>NEC Laboratories Europe - Network Research Division NEC Europe Limited
>Registered Office: NEC House, 1 Victoria Road, London W3 6BL
>Registered in England 283
>_______________________________________________
>alto mailing list
>[email protected]
>https://www.ietf.org/mailman/listinfo/alto


_______________________________________________
alto mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/alto

Reply via email to