-
Notifications
You must be signed in to change notification settings - Fork 10
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Modelling of Transponders #29
Comments
These are the current models: WSON topology (https://tools.ietf.org/html/draft-ietf-ccamp-wson-yang-23):
Flexi-grid topology (https://tools.ietf.org/html/draft-ietf-ccamp-flexigrid-yang-05):
Optical-impairments topology (https://tools.ietf.org/html/draft-ietf-ccamp-optical-impairment-topology-yang-02):
|
See also open issues #5; ietf-ccamp-wg/ietf-ccamp-layer0-types-ext-RFC9093-bis#47; #12 and ietf-ccamp-wg/ietf-ccamp-layer0-types-ext-RFC9093-bis#51 |
For the optical impairment aware L0 topology, the modelling of the following optical transponder properties are needed:
|
To complete the picture herewith the parameters from draft-ietf-ccamp-dwdm-if-param-yang-03:
|
Some initial questions/doubts considering for example a 400G transponder. Is it possible that the number of OTSis supported depends on the operational model. For example. the transport can take an OTUC4 and transport it either over a 400G OTSi or over 4x 100G OTSis. When the transponder is capable to support 4x 100G OTSis, can we assume that it is always capable to put all of them into a single media channel (if available) or should we describe the capability of the transponder to put multiple OTSis into one media channel? The model allows for different OTSis to use different operation modes but we expect that in general the same operational model is used. Would it be a a realistic use case to consider cases where the transponder has flexibility also on the client signal configuration (for example it is capable to have as input either an OTUC4 or 4x OTUC1/OTU4)? |
SB> My understanding of the Dieter's proposal is that the last attrubute in the list
SB> see above
SB> Why the fact to use 1 or more media channel should impact topology model for optical impairments? It seems to me more a problem of path setup .
SB> During last call it was said that it seems reasonable to consider that OTSi belongong to the same OTiG can use the same operational mode.
|
[Italo] My doubt is that in this case the server reports that the OTSiG:OTSi cardinality is 4 (since it is possible to carry an OTUC4 over 4x OTSi@100G) and that the OTSi supports a 400G operating mode but it does not report that these two capabilities are mutually exclusive since the transponder cannot be configured to carry an OTUC16 over 4xOTSis@400G.
[Italo] I agree that it is a problem of path setup/computation but I think it should be addressed otherwise path computation finds a path which is feasible from optical impairments perspective but unfeasible because of transponder capabilities.
[Italo] My doubt is not about the possibility to use the same operational model but about the possibility for an implementation to require this. If an implementation requires all the OTSis within the same OTSiG group to use the same operational model for example it would not be possible to carry an OTUC4 over one OTSi@200G and 2xOTSis@100G even if both operational modes are supported. If path computation is not aware of this transponder limitation it could find an optical path which seems feasible but it is actually unfeasible because of transponder capabilities. As above it is not an optical impairment issue but a path setup/computation issue.
|
I have double checked the current proposal from Dieter with existing definitions in WSON and Flexi-grid topology and found that these attributes are missing but that requires further discussion/investigation:
I am not sure whether the FEC type is implicitly defined by the operational mode or not. In other words, is it possible to configure different FEC types for the same operational mode?
The definition and use of this attribute is not fully clear to me but at the first glance is seems more related to the 3R and/or client signal adaptation capabilities
The definition and use of this attribute is not fully clear
I think we can replace this attribute with these two attributes proposed by Dieter which provides more information about the tuning capabilities:
It looks like the same as the max. OTSiG:OTSi cardinality attribute proposed by Dieter. I think it is applicable to both WSON and flexi-grid networks.
I am not sure we really need this attribute. The LLC should provide enough information (label-restrictions) to describe any limitation the transponder has on the media-channel frequency-slot(s) it could use. This information includes also the type of grid the transponder can use. |
I have also noted that the optical-impairments topology allows one TTP to have multiple transponders (transponders-list* [transponder-id]) but each entry of the the transponder-list defines the capability of one OTSi Also the DWDM interface model seems describing a transponder with supports only one OTSi Am I missing something? |
[Dieter] The max OTSiG:OTSi cardinality is one constraint imposed by the OT HW. A max OTSiG data rate in terms of n x 100G is an additional constraint needed for path computation in case inverse muxing is supported. This is not directly related to L0 and optical impairments but should probably be added.
[Dieter] I don't think that putting the OTSi signals into one or more media channels is an issue of the optical transponder capabilities.
[Dieter] As discussed last week, there was agreement that the OTSi signals belonging to the same OTSiG are usually configured homogeneously. However, the YANG model should not prohibit that and shall allow to use different OTSi configurations. If the data rate for the different OTSi signals are different, the OTSiG must be capable to subdivide the OTSiG data accordingly. This will add complexity to the path computation function because the number of potential configurations increases.
|
[Dieter] This needs to be reviewed and we have to check the TTP definition carefully. The OTSi termination function terminates the photonic signal and the OTSiG termination function performs the is dis-aggregation/aggregation of the OTSiG client signal into multiple OTSi flows. |
Related to optical transponder capabilities , we have already a specif issue and comment form Esther |
It may be useful to distinguish the impairment model from routing needs. For the impairment model it is not important if two OTSi carry the same OTUCn or not. For calculating the impairment it would suffice to know the signal itself and the spectral distance to neighbor channels. So there is a need to identify the co-routing requirement for OTSi, let's call it OTSI-CRG for the time being (members of an OTSIG can be individually routed) To calculate the feasibility of an OTSi-CRG, one would need to definefor member-OTSi the minimum distance to an OTSi of the same CRG and the additional distance to an OTSi from a different CRG. That would allow
|
Related to the discussion had last week and AP to restructure the model taking into account Esther's comment |
Presentation1.pptx |
Here is my thought of the naming of operational modes
Then where should I put the information that this transceiver is a “foo” transceiver? Then how can I avoid wrong implementations such as
Which does not make sense because this is the list of a given transceiver capability (foo or bar, not both)
|
If I got correctly your point and suggestion, the better encoding would be simple to add, in addition to "id" of transceiver also a new leaf indicating the "type" of transceiver and then the related list of possible "modes" supported . Correct? |
@EstherLerouzic : I am not sure whether you are assuming that a transponder type « foo » cannot interoperate with a transponder type « bar », even if they are from the same vendor or not. If it is possible that the two transponder types can interoperate, how can the controller know that e..g, mode 1 for transponder type « foo » can interoperate with mode 2 for transponder type « bar »? |
@ggrammel : Two questions for clarification based on your slide 4
|
Yes ! |
I have made no assumptions on interoperability: only wanted to highlight foo capability in terms of modes.
] |
This goes into the same direction as our last discussion about operational modes for interfaces vs. topology. The outcome there was that we need two different identifiers because there is a 1:n relationship between both. So one Identifier that encodes the interoperability (application codes and organizational codes) and another for the interface configuration setting (if-mode). Hint: https://tools.ietf.org/html/draft-ietf-ccamp-dwdm-if-lmp-02 says: IOW it allows organizations to encode more tha just the organization in the string. |
The slide wasn't meant to provide a particular semantic for "Application Identifier". We know that the term "Application Code" is reserved for Standardized codes, and for non-standardized codes the term "Application Identifier" is used. So on p.4 I used "Application Identifier" since it appears to be more general.
Yes, this was thanks to Sergio's observation that the if-mode describes only a transceiver and there is a 1:n relationship between if-mode and mode (used by topology-draft). The current dwdm-if YANG model has indeed only a single string that could be used for that purpose. So for the Interface the following may work. Note that in order to distinguish between different semantics in "modes" I added "if-mode" to indicate this is different from "mode" as used by the current topology draft. The application-identifiers-list may contain application-codes as well as organizational-codes. It's purpose is to trace back applications to if-modes and the dwdm-if model doesn't require a particular semantic.
@EstherLerouzic The proposal is indirectly addressing also the issue you brought up. When connecting two transceivers in the topology model, it is based on application identifiers, but upon selecting the application identifier foo on node A would be configured with its associated if-mode-A and bar on node Z would get its if-mode-Z. |
@ggrammel @dieterbeller : Sergio and I have tried to better understand your viewpoint and prepared few slides to describe the two approaches. To simplify the discussion, let's consider only standard application codes at this stage. We understand that slide 1 represents the OpenConfig approach, where, the mapping between the application code, selected by path computation, and the if-mode is done internally by the transponder device, while slide 2 represents your proposal where this mapping is done by the domain controller. Is our understanding correct? |
@italobusi & @sergiobelotti
added a 3rd slide to clarify my interpretation. Note that the selection is a little different. |
For the time being, let's assume that A, B, C and D are application codes only
Yes, the intention is to show that the mapping between if-mode and application code is internal to the transponder device and not exposed to the domain controller
Yes, if-mode
It was text to be rephrased Slides updated to fix these issues:
Few questions of mine about slide 3:
|
I'm a bit confuse of your added slide: our thought was that you have "interface" set of parameters configured corresponding to an if-mode, covering more than 1 (1:N) possible application. So I do not understand the meaning of your slide with respect the other , in particular slide 2 |
Additional comments/text proposals: |
Wondering if there is some misunderstanding to clarify. Looking at the table, it references for the Organizational mode that a set of explicit parameters in column-2 would be "explicit". Is it a typo? My understanding is:
|
@ggrammel : I share your understanding of model 1B but I am not sure I have understood your initial point about misunderstanding in the table. I assume you are referring to the table proposed in https://github.com/ietf-ccamp-wg/draft-ietf-ccamp-optical-impairment-topology-yang/files/5424280/Transponder-text-section-2.5_elr.docx My understanding is that:
In other words, it seems that for the attributes in column-2, ITU-T application codes and organizational modes have adopted a different approach. The YANG model and the table seem align with this understanding. |
#29 (comment) was setting out that e.g. Min/Max frequency and tunability grid are part of the definition in e.g. OpenROADM and is in line with the G.698.2 table. Parameters like this do not show up in Model-1B outside for the Explicit mode. It is confusing to see those popping up at a location where they were intended to be hidden in the first place. |
@ggrammel : Yes, your understanding of the model 1B is correct.
@italobusi : Yes, you got correctly . The only parameters that also organizational mode is showing up is the range of soem attributes like rx-power-channel or tx-power-channel as we discussed and agreed in #29 (comment)
@ggrammel : correct, this is also my understanding of organizational mode, the parameters I've indicated above are not implicit in the organizational mode.
|
This is still confusing. It appears the change in the model came with the merge 11 days ago, but it is not really clear to me what caused this change. In any case the Text seems to contradict itself and needs to be cleaned up. |
|
seems the references are all a bit coarse, let me try to describe the location in addition to the links, so that it is easier to find, it's a bit hard to summarize everything:
So when comparing it to definition of Application Code, my reading is that apart from the organization, Application Code and organizational mode are considered the same.
What the discussion highlighted is that there is a discrepancy in the understanding about what an organizational code describes and we need to get this straight. Let me try to describe the issue it in other words: IF the set of explicit parameters like "min/max frequency" would be required to allow interoperability with Operational modes, THEN the following condition of the description of Organizational mode is not met:
|
Meanwhile, I found some time for drafting some text describing the Organizational Mode option re transceiver capabilities: Organizational Mode: The major reason for these transceiver presets, is that the attribute values typically cannot be configured independently and are therefore advertised as supported operational mode capabilities. |
no issue with this text |
2020-10-29 AI: @dieterbeller : Provide complete list of desired parameters to add to the operational mode definition |
May I suggest to add a couple of lines to specify the ratio of having a subset of parameters explicitly reported?
Something like this
“A subset of the transceiver properties subsumed by the operational mode may be explicitly exposed; this allows transceivers of different performances interoperate without the need to introduce a specific operational mode. One use case includes transceivers which differ only by the input power sensitivities. Another use case include transceivers which differ only by the tunability range of the carrier frequency. The set of transceiver characteristic explicitly exposed is limited to input and output power and tunability capabilities of the transceiver.”
|
@egriseri the use case described had been discussed when taking the decision for Model 1B. The Explicit mode allows to define parameters such as different receiver sensitivity or tunability ranges. In particular, https://github.com/ietf-ccamp-wg/draft-ietf-ccamp-optical-impairment-topology-yang/files/5104243/if-mode-clarification_options-200820-github.pptx introduced a The Issues I have with the approach to add explicit parameters to extend the operational mode are:
In summary: parameters defining an organizational mode should stay implicit to support simple matching conditions. Where extensions are required, an explicit mode with reference to the organizational mode appears to do the job. |
Based on Enrico's comment #29 (comment), I updated the text above (see #29 (comment)) as follows: Organizational Mode: The major reason for these transceiver presets is the fact that the attribute values typically cannot be configured independently and are therefore advertised as supported operational mode capabilities. |
@dieterbeller: the new text would need discussion. While there are always differences between implementations and samples, they are supposed to be considered by the MSA-organization in such a way that they interoperate. Suggestion is to first nail down the requirements before jumping in a solution. |
@ggrammel : the intent of the transponder model part is exactly to "model" not invent , the existing way to determine optical signal compatibility. This why we have described in YANG with the triplet (standard, organizational and explicit).
@italobusi @EstherLerouzic : Taking into account the new text from Dieter #29 (comment) |
@sergiobelotti<mailto:notifications@github.com> can you be more specific about “organizational mode, how it works”? In past discussions @Dieter claimed it would be a simple match condition in OpenConfig. If the purpose of the operational mode is to mimic how OpenConfig works, then what needs to change?
The model there looks like this: https://github.com/openconfig/public/blob/master/release/models/optical-transport/openconfig-terminal-device.yang
leaf operational-mode {
type uint16;
description
"Vendor-specific mode identifier -- sets the operational
mode for the channel. The specified operational mode must
exist in the list of supported operational modes supplied
by the device";
//
// Ideally, this leaf should be a leafref to the supported
// operational modes, but YANG 1.0 does not allow a r/w
// leaf to be a leafref to a r/o leaf.
}
From: sergiobelotti <notifications@github.com>
Reply-To: ietf-ccamp-wg/draft-ietf-ccamp-optical-impairment-topology-yang <reply@reply.github.com>
Date: Friday, October 30, 2020 at 16:09
To: ietf-ccamp-wg/draft-ietf-ccamp-optical-impairment-topology-yang <draft-ietf-ccamp-optical-impairment-topology-yang@noreply.github.com>
Cc: Gert Grammel <ggrammel@juniper.net>, Mention <mention@noreply.github.com>
Subject: Re: [ietf-ccamp-wg/draft-ietf-ccamp-optical-impairment-topology-yang] Modelling of Transponders (#29)
[External Email. Be cautious of content]
@ggrammel<https://github.com/ggrammel> : the intent of the transponder model part is exactly to "model" not invent , the existing way to determine optical signal compatibility. This why we have described in YANG with the triplet (standard, organizational and explicit).
Text proposed perfectly describes what it is the sate of art of organizational mode, how it works, no alternative on that, and obviously who has implementation on that knows very well this. We do not have need of requirements to describe what is already there , in field.
The text you mention, is "my" proposal that obviously has been completed and improved by Dieter and Enrico suggestion.
I've already updated text of section 2.5 along these lines and uploaded
here is a change proposal for the text.
Transponder-text-section-2.5_elr.docx<https://github.com/ietf-ccamp-wg/draft-ietf-ccamp-optical-impairment-topology-yang/files/5424280/Transponder-text-section-2.5_elr.docx>
I added a table with my understanding of the explicit/implicit presence of parameters
@EstherLerouzic<https://github.com/EstherLerouzic> : I'm generally fine with the table . One comment: you wrote
"
-maximum channel
power on receiver
-max total power on receiver
Are not the same ??? In the model we have "rx-total-power-max" .
About your comment " I thought that there was a leaf ref to the configured mode, no ? and maybe the instance related settings such as current optical emitter power, frequency, ..."
Yes , it is exactly what you describe. I will add text along your lines.
Herewith attached the new proposed text.
Transponder-text-section-2.5_elr-sb-261020.docx<https://github.com/ietf-ccamp-wg/draft-ietf-ccamp-optical-impairment-topology-yang/files/5438399/Transponder-text-section-2.5_elr-sb-261020.docx>
Additional comments/text proposals:
Transponder-text-section-2.5_elr-sb-261020-ib.docx<https://github.com/ietf-ccamp-wg/draft-ietf-ccamp-optical-impairment-topology-yang/files/5440306/Transponder-text-section-2.5_elr-sb-261020-ib.docx>
@italobusi<https://github.com/italobusi> @EstherLerouzic<https://github.com/EstherLerouzic> : Taking into account the new text from Dieter ietf-ccamp-wg/draft-ietf-ccamp-optical-impairment-topology-yang#29 (comment)<#29 (comment)>
I've uploaded the new section 2.5 to be used for uploading of the new version of the draft in the view of IETF 109. Remain open some comment about definitions from Italo but we can survive with the present text for uploading and take the action to verify with our internal Q6 expert for better definition of transponder/transceiver.
Transponder-text-section-2.5_301020.docx<https://github.com/ietf-ccamp-wg/draft-ietf-ccamp-optical-impairment-topology-yang/files/5465887/Transponder-text-section-2.5_301020.docx>
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub<#29 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ADTQOVFPMA4TRGDBO3SLHFLSNLJIVANCNFSM4LC6BAFA>.
Juniper Business Use Only
|
This discussion with multiple embedded comments is quite hard to follow :)
I just copied the list of bullet points that were on the text and wrapped them a bit... So I may have misunderstods these two bullet points (listed in the docx file). here is my understanding: I have read https://github.com/ietf-ccamp-wg/draft-ietf-ccamp-optical-impairment-topology-yang/files/5465887/Transponder-text-section-2.5_301020.docx and have no comment for the text, except that maybe we need to better name "supported receiver power range" and "supported maximum total power" thanks ! |
Your intent is clear, but is not what I understood for the organizational mode.
This is not my understanding. The interoperability is done by matching of (organization-identifier, operational-mode). Letting optical power and carrier frequency related properties avoid un-necessary duplication of operational modes in case sligthly different optical front end are used.
The explicit parameters are listed in @dieterbeller answer (and in the agreed Yang code). No more.
See answers above. |
@egriseri<mailto:notifications@github.com>
1. what is your understanding of an Operational Mode? Is this understanding in line with OpenConfig definitions or any other organization? Would be great to align here.
2. If the interoperability is done by matching of (organization-identifier, operational-mode), then there is no need to check anything. However when you add a frequency range to organizational modes and the frequency range of transmitter and receiver are different, then there is no interoperability on the non-overlapping range. This complicates the path calculation quite a bit.
3. Thanks for confirming that the parameters are exactly the ones in discussion and are not going to be extended.
4. What are your thoughts about how to circumvent the incompatibility issue?
From: egriseri <notifications@github.com>
Reply-To: ietf-ccamp-wg/draft-ietf-ccamp-optical-impairment-topology-yang <reply@reply.github.com>
Date: Friday, October 30, 2020 at 19:32
To: ietf-ccamp-wg/draft-ietf-ccamp-optical-impairment-topology-yang <draft-ietf-ccamp-optical-impairment-topology-yang@noreply.github.com>
Cc: Gert Grammel <ggrammel@juniper.net>, Mention <mention@noreply.github.com>
Subject: Re: [ietf-ccamp-wg/draft-ietf-ccamp-optical-impairment-topology-yang] Modelling of Transponders (#29)
[External Email. Be cautious of content]
@ggrammel<https://github.com/ggrammel>
The desire was to keep the Application code and organizational mode as simple matching condition, while detailed parameters were pushed to the explicit mode where parameter checks need to apply naturally.
Your intent is clear, but is not what I understood for the organizational mode.
The Issues I have with the approach to add explicit parameters to extend the operational mode are:
1. a compatibility check practically becomes the same as the compatibility check required for the explicit mode: all parameters and ranges need to fit. The intention however was to use it as a sole matching criteria.
This is not my understanding. The interoperability is done by matching of (organization-identifier, operational-mode). Letting optical power and carrier frequency related properties avoid un-necessary duplication of operational modes in case sligthly different optical front end are used.
1. What are the limits for adding additional parameters to further extend the operational mode? Adding many explicit parameters making it practically the same as the explicit mode.
The explicit parameters are listed in @dieterbeller<https://github.com/dieterbeller> answer (and in the agreed Yang code). No more.
1. The explicit mode has a leafref (described above) to the organizational mode. By which mechanism can we exclude an inconsistency between explicit parameters defined in the explicit mode referencing an operational mode and explicit parameters defined there?
In summary: parameters defining an organizational mode should stay implicit to support simple matching conditions. Where extensions are required, an explicit mode with reference to the organizational mode appears to do the job.
See answers above.
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub<#29 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ADTQOVAF333TJFILE23J3N3SNMBEJANCNFSM4LC6BAFA>.
Juniper Business Use Only
|
My personal understanding of Operational Mode is mainly to avoid un-necessary duplication of Organizational Modes in case many vendors (or the same vendor) have interoperable optics with small differences in power range and tunability. This happens even within the same vendor portfolio, thus why impose a duplication? In OpenConfig, I read that the optical channel in OpenConfig is configured by the (carrier) frequency, target-output-power, as well as by an operational mode.
I'm sorry I disagree: in a DWDM system power and frequency ranges must be in any case be checked against the DWDM system settings and capabilities. This is always required no matter the Application Code, Organizational Mode or Explicit Mode is used.
You're welcome. The rationale is to collect under the Operational Mode "what really matters" for interoperability.
Don't we have the same issue when an Application Code is referenced to from within an Explicit Mode? |
YANG model and tree view aligned with ietf-optical-impairment-topology.yang and ietf-optical-impairment-topology.tree on GitHub, Section "2.5 Transpomders" updated based on Open Issue #29
@egriseri trying to summarize in a copy/paste session from your earlier comment to make it more readable: 1a) My personal understanding of Operational Mode is mainly to avoid un-necessary duplication of Organizational Modes
|
closed issue with commit 8b7c07e |
There are three models for transponders in WSON topology, Flexi-grid topology and optical-impairment topology
It is suggested to reconcile these models
The text was updated successfully, but these errors were encountered: