Hi Olivier,
Many thanks for your great review. Please see my comments inline (noted <BV>).
Changes are included:
https://protect2.fireeye.com/v1/url?k=dee2e172-8179d836-dee2a1e9-861d4
1abace8-9257af7700956e47&q=1&e=c6fd3bf1-3566-493c-be5c-8b64514e8a3e&u=
https%3A%2F%2Fgithub.com%2Fdetnet-wg%2Fdraft-ietf-detnet-flow-informat
ion-model%2Fcommit%2F5aed47ec5b0215cf4b6ac16d67271fe7387f7179%23diff-6
49f76b4565f6dacda6b86e0092367053c96a461fb5b114a62facdde502db7e7
Please, let us know if you intend to finetune the resolution of your comments.
Thanks & Cheers
Bala'zs
-----Original Message-----
From: Olivier Bonaventure via Datatracker <noreply@xxxxxxxx>
Sent: Friday, November 6, 2020 9:56 PM
To: tsv-art@xxxxxxxx
Cc: detnet@xxxxxxxx;
draft-ietf-detnet-flow-information-model.all@xxxxxxxx;
last-call@xxxxxxxx
Subject: Tsvart last call review of
draft-ietf-detnet-flow-information-model-11
Reviewer: Olivier Bonaventure
Review result: Almost Ready
This document has been reviewed as part of the transport area review team's ongoing effort to review key IETF documents. These comments were written primarily for the transport area directors, but are copied to the document's authors and WG to allow them to address any issues raised and also to the IETF discussion list for information.
When done at the time of IETF Last Call, the authors should consider this review as part of the last-call comments they receive. Please always CC tsv-art@xxxxxxxx if you reply to or forward this review.
This review was requested in a very short period of time and I could not read all the Detnet related materials. I focus on issues that could be relevant for the transport area.
The introduction indicates that the Detnet architecture supports IP and MPLS flows, but section 5.4.2 indicates that and IP flow can be specified using the following attributes :
a. SourceIpAddress
b. DestinationIpAddress
c. IPv6FlowLabel
d. Dscp (attribute)
e. Protocol
f. SourcePort
g. DestinationPort
h. IPSecSpi
I would personally qualify the flows that include transport layer information as layer-4 and not IP flows. Maybe a note in the introduction should be mentioned. In the document, it is unclear to me whether some of these attributes are optional or can be specified as wild cards. Maybe this is described in another document.
<BV> Yes, these attributes are defined in [I-D.ietf-detnet-ip] and they can be specified as wild cards. For a given network scenario some of them or all of them may be used. In case of using e, f, and g, for DetNet IP Flow Identification then as you wrote it can be treated as layer-4 flow.
In section 5.4.2. the sentence before the attribute list points explicitly to [I-D.ietf-detnet-ip]. I have added a note at the end of the section.
<NEW Text>
Using wild cards for these attributes are specified in [I-D.ietf-detnet-ip].
<END>
The information model defines various attributes. Some of the definitions are very precise and mention the measurement unit. For others, this is less clear.
For example, intervals are defined in units of nanoseconds, but payloadsize is not defined as a number of bytes.
<BV> Right. I have added new text to clarify this in section 5.5 in the paragraph after the list of those attributes.
<OLD Text>
These attributes can be used to describe any type of traffic (e.g.,
CBR, VBR, etc.) and can be used during resource allocation to
represent worst case scenarios.
<NEW Text>
These attributes can be used to describe any type of traffic (e.g.,
CBR, VBR, etc.) and can be used during resource allocation to
represent worst case scenarios. Interval is specified as an integer
number of nanoseconds. MaxPayloadSize is specified in octets
per second.
<END>
Some information elements appear unclear to an external observer like me:
5.9.4. Maximum Loss of the DetNet Flow
MaxLoss defines the maximum Packet Loss Ratio (PLR) requirement for
the DetNet flow between the Ingress and Egress(es).
There is no measurement interval define for the packet loss ratio.
<BV> Right. Proposed correction.
<OLD Text>
MaxLoss defines the maximum Packet Loss Ratio (PLR) requirement for
the DetNet flow between the Ingress and Egress(es).
<NEW Text>
MaxLoss defines the maximum Packet Loss Ratio (PLR) requirement for
the DetNet flow between the Ingress and Egress(es) and the loss
measurement interval.
<END>
5.9.6. Maximum Misordering Tolerance of the DetNet Flow
The definition is unclear and does not make it easy to measure this
value
<BV> Right. Current text add some hints for measurement, but added some more clarification. Proposed correction.
<OLD Text>
MaxMisordering describes the tolerable maximum number of packets that
can be received out of order. The maximum allowed misordering can be
measured for example based on sequence number. The value zero for
the maximum allowed misordering indicates that in order delivery is
required, misordering cannot be tolerated.
<NEW Text>
MaxMisordering describes the tolerable maximum number of packets
that can be received out of order. The value zero for the maximum
allowed misordering indicates that in order delivery is required,
misordering cannot be tolerated.
The maximum allowed misordering can be measured for example based
on sequence number. The difference of sequence number values in
consecutive packets at the Egress cannot be bigger than
"MaxMisordering + 1".
<END>
6.3.1. Minimum Bandwidth of the DetNet Service
>From the definition, it is unclear to me whether the payload only is
used to
compute the bandwidth or whether this includes the headers and if so
which
<BV> Right. It excludes additional DetNet header (if any). Proposed clarification.
<OLD Text>
MinBandwidth is the minimum bandwidth that has to be guaranteed for
the DetNet service. MinBandwidth is specified in octets per second.
<NEW Text>
MinBandwidth is the minimum bandwidth that has to be guaranteed for
the DetNet service. MinBandwidth is specified in octets per second and
excludes additional DetNet header (if any).
<END>
I was surprised by the following paragraph
6.4. Connectivity Type of the DetNet Service
Two connectivity types are distinguished: point-to-point (p2p) and
point-to-multipoint (p2mp). Connectivity type p2mp is created by a
transport layer function (e.g., p2mp LSP). (Note: mp2mp connectivity
is a superposition of p2mp connections.)
Does the note precludes the utilisation of multicast to support the detnet service ? If so, I don't think that this restriction should be part of an information model. It should be specified in an architecture document.
<BV> Right, text is rather misleading and not accurate enough. Multicast is allowed. Proposed clarification.
<OLD Text>
Two connectivity types are distinguished: point-to-point (p2p) and
point-to-multipoint (p2mp). Connectivity type p2mp is created by a
transport layer function (e.g., p2mp LSP). (Note: mp2mp connectivity
is a superposition of p2mp connections.) <NEW Text>
Two connectivity types are distinguished: point-to-point (p2p) and
point-to-multipoint (p2mp). Connectivity type p2mp may be created by a
forwarding function (e.g., p2mp LSP). (Note: from service perspective
mp2mp connectivity can be treated as a superposition of p2mp
connections.)
<END>
6.6. Rank of the DetNet Service
The DnServiceRank attribute provides the rank of a service instance
relative to other services in the DetNet domain. DnServiceRank
(range: 0-255) is used by the network in case of network resource
limitation scenarios.
>From this description I do not know whether a high value rank gets more
resource than a low value one. I would suggest to clarify this in the definition of the rank.
<BV> Right. :--))) Proposed clarification. Sentence added at the end of the paragraph.
<OLD Text>
The DnServiceRank attribute provides the rank of a service instance
relative to other services in the DetNet domain. DnServiceRank
(range: 0-255) is used by the network in case of network resource
limitation scenarios.
<NEW Text>
The DnServiceRank attribute provides the rank of a service instance
relative to other services in the DetNet domain. DnServiceRank
(range: 0-255) is used by the network in case of network resource
limitation scenarios. High value rank gets preferred over a low value
one.
<END>
