Re: [Dots] AD review of draft-ietf-dots-use-cases-17

[Daniel Migault <[e-mail]>]

Hi,

Thank you for the review. Please find my response in line as well as on the
git repo [1].  WG, co-authors, please review section 3.3 by Friday EOB.

In summary we addressed all comments, major changes are:

a) Figure 2 with a channel from DOTS client and DOTS server with th
eaddition of the following text:
"""
In some cases the communication between the enterprise DOTS client and
the DOTS server of the DDoS Mitigation Service Provider may go through
the ITP carrying the DDoS attack, which would affect the
communication. On the other hand, the communication between the DOTS
client and DOTS server may take a path that is not undergoing a DDoS
attack.
"""

b) telemetry

We clarified DOTS client/server involved.

c) it was unclear to me how to address the following comment.


>    The communication between a network administrator and the
>    orchestrator is also performed using DOTS.  The network administrator
>    via its web interfaces implements a DOTS client, while the
>    Orchestrator implements a DOTS server.
>
> nit: as written, this is saying that the network administrator has a
> web interface.  I think "its" is supposed to refer to something else.
>
> <mglt>
What we are trying to say is that the network administrator sees its web
interface, and instruct the DOTS client from that interface. I have not
made any change to address that concern, as I do not clearly see what is
confusing.
</mglt>


Yours,
Daniel

[1]
https://github.com/dotswg/dots-use-cases/commit/e251fb8abb51ba0c68471e847037daf2e81d38aa


On Tue, Jul 2, 2019 at 6:37 PM Benjamin Kaduk <[e-mail]> wrote:

> First off, a few housekeeping items:
>
> (1) This document lists seven authors, and per RFC 7322 I/the IESG needs
> to specially consider this and essentially make an exception to have
> more than five authors.  Can you please confirm that all listed authors
> have made substantial contributions, so that I can take that to the IESG
> and get it approved?
>
> (2) The shepherd writeup indicates that three authors (Stefan, Bob, and
> Nik) have not indicated conformance with BCPs 78 and 79.  I don't think
> I can issue the IETF LC until that gets straightened out, so please
> confirm that we're all squared away!
>
> (3) Recently the IESG has been trying to exert some gentle backpressure
> against publishing Informational use-cases/requirements drafts, when they
> serve only as input to future protocol specifications and do not have
> lasting archival value on their own.  I do see in the shepherd writeup that
> the working group did reach consensus to publish this document and think
> there's enough value in it to be worth publishing; I just mention this so
> that people aren't surprised if the IESG evaluation comes back with
> questions about whether we should be publishing the document at all.
>
> Other than those, the document is generally in good shape; there's just
> a few substantive questions buried in the editorial nits.
>
> On to the section-by-section comments:
>
> Section 2
>
>    o  DDoS Mitigation Service: designates a DDoS mitigation service
>       provided to a customer and which is scoped to mitigate DDoS
>       attacks.  [...]
>
> I don't really think that using the lowercase-'s' version to define the
> uppercase-'S' version of the term is going to help anyone.
>
<mglt>
The text has been replaced by the following:

* DDoS Mitigation Service: designates a service provided to a
customer to mitigate DDoS attacks.  Services usually involve Service
Level Agreement (SLA) that have to be met.  It is the responsibility of
the DDoS Service provider to instantiate the DDoS Mitigation System to
meet these SLAs.

</mglt>

>
> Section 3.1
>
> It's a little surprising that we have the two bullet points near the top
> about the enterprise DMS acting as a DOTS client for the first kind of
> service but as a DOTS server for the second kind, but then we never seem
> to talk about that second kind of service again in the document.
> Perhaps we should just explicitly say that it's similar to the first
> kind and not covered further?
>
> <mglt>
I agree that could be mentioned explicitly. Here is the proposed text to
address that concern.

 The two scenarios, thought different, have similar interactions between
the DOTS client and server. For the sake of simplicity, only the first
scenario will be detailed in this section.
</mglt>

   When the enterprise DMS detects an inbound DDoS attack targeting its
>    resources ( e.g. servers, hosts or applications), it immediately
>    begins a DDoS Mitigation.
>
> I'd consider clarifying that this mitigation is entirely local within
> the enterprise, so that contacting the ITP in the next step is a clear
> escalation.
>
> <mglt>
To address the concern we specify that the mitigation is handled locally as
well as we explicitly indicate the escalation procedure.  I believe the
following text address your concern:

When the enterprise DMS locally detects an inbound DDoS attack targeting
its resources ( e.g. servers, hosts or applications), it immediately
begins a DDoS Mitigation.

During the course of the attack, the inbound traffic volume exceeds the
50% threshold and the enterprise DMS escalates the DDoS mitigation.  The
enterprise DMS DOTS client signals the DOTS server on the upstream ITP
to initiate DDoS Mitigation. The DOTS server signals the DOTS client
that it can serve this request, and mitigation is initiated on the ITP
network by the ITP DMS.
</mglt>


>    related information.  Once the DDoS attack has ended, or decreased to
>    the certain level that the DOTS client can handle by itself, the DOTS
>    server signals the enterprise DMS DOTS client that the attack has
>    subsided.
>
> I think it's the enterprise DMS that is handling the attack, not the
> DOTS client directly...
>
<mglt>
This is correct. This has been corrected as follows:

Over the course of the attack, the DOTS server of the ITP periodically
informs the DOTS client on the enterprise DMS mitigation status,
statistics related to DDoS attack traffic mitigation, and related
information. Once the DDoS attack has ended, or decreased to the certain
level that the enterprise DMS can handle by itself, the DOTS server
signals the enterprise DMS DOTS client that the attack has subsided.
</mglt>

>

   The enterprise DMS then requests the ITP to terminate the DDoS
>    Mitigation.  The DOTS server on the ITP receives this request and
>
> .... but this one is the DOTS client.
>
> <mglt>

Yes we sometime probably abused metonymy, but I agree the more specific
the better. I believe the following text is clarifying.

The DOTS client on the enterprise DMS then requests the ITP to terminate
the DDoS Mitigation. The DOTS server on the ITP receives this request
and once the mitigation has ended, confirms the end of upstream DDoS
Mitigation to the enterprise DMS DOTS client.

</mglt>

>    o  (a) A DDoS attack is initiated against resources of a network
>       organization which has deployed a DOTS-capable DMS - typically a
>       DOTS client.
>
> We probably want to reiterate in a parenthetical "network organization
> (here, the enterprise)" the terminology we're using.
>
> <mglt>
Here is the current text:

* (a) A DDoS attack is initiated against resources of a
network organization (here, the enterprise) which has deployed a
DOTS-capable DMS - typically a DOTS client.
</mglt>


>    o  (d) The DOTS server which receives the DOTS Mitigation request
>       determines that it has been configured to honor requests from the
>       requesting DOTS client, and honored its DDoS Mitigation by
>       orchestrating its DMS.
>
> nit: I think s/honored/honors/ to stay in the present tense.
>
> <mglt>
I agree. here is the proposed text:

* (d) The DOTS server which receives the DOTS Mitigation request
determines that it has been configured to honor requests from the
requesting DOTS client, and honors its DDoS Mitigation by orchestrating
its DMS.

</mglt>


>    o  (e) While the DDoS Mitigation is active, DOTS server regularly
>       transmits DOTS DDoS Mitigation status updates to the DOTS client.
>
> nit: "the DOTS server" or "servers regularly transmit".
>
> <mglt>
I agree, here is the corrected text:

* (e) While the DDoS Mitigation is active, the DOTS server
regularly transmits DOTS DDoS Mitigation status updates to the DOTS
client.
</mglt>


> Section 3.2
>
>    As such, this use case likely to match large enterprises or large
>    data centers, but not exclusively.  [...]
>
> nit: "is likely"
>
> <mglt>
This has been added as follows:

As such, this use case is
likely to match large enterprises or large data centers, but not
exclusively.
</mglt>

   In this scenario, an Enterprise Network has entered into a pre-
>    arranged DDoS mitigation assistance agreement with one or more other
>    DDoS Mitigation Service Providers in order to ensure that sufficient
>    DDoS mitigation capacity and/or capabilities may be activated in the
>    event that a given DDoS attack threatens to overwhelm the ability of
>    a given DMS to mitigate the attack on its own.
>
> We could perhaps say "overwhelm the ability of the enterprise's or any
> other given DMS" since in most cases the enterprise DMS is the one at
> risk of first being overwhelmed.
>
> <mglt>
I agree that is better. Here is the modified text:

In this scenario, an Enterprise Network has entered into a pre-arranged
DDoS mitigation assistance agreement with one or more other DDoS
Mitigation Service Providers in order to ensure that sufficient DDoS
mitigation capacity and/or capabilities may be activated in the event
that a given DDoS attack threatens to overwhelm the ability of the
enterprise's or any other given DMS to mitigate the attack on its own.
</mglt>


> Is the fact that the C<-->S DOTS traffic does not go through the ITP in
> Figure 3 an intentional change from Figure 2 (in that they are expected
> to be communicating "out of band" or not through the enterprise's normal
> transit)?  Some readers might see this and get confused if this
> communication is still supposed to be going along the regular transit
> path.
>
> <mglt>
The intention was to indicate the DDoS Mitigation Service Provider is not
the upstream ITP and thus communication MAY have to transit through the
ITP.  I understand from your comment that it might be interpreted as the
communication MUST go through the ITP.

I propose to have the DMS communicating directly via a distinct channel and
mention explicitly that the DOTS channel MAY transit via the ITP or may use
a dedicated path.

Here is the updated figure and the additional text:

       +------------------+        +------------------+
       | Enterprise       |        | Upstream         |
       | Network          |        | Internet Transit |
       |                  |        | Provider         |
       |      +--------+  |        |             DDoS Attack
       |      | DDoS   |  | <=================================
       |      | Target |  | <=================================
       |      +--------+  |        |                  |
       |                  |        |                  |
       |                  |        +------------------+
       |                  |
       |                  |        +------------------+
       |                  |        | DDoS Mitigation  |
       |                  |        | Service Provider |
       |                  |        |                  |
       |  +------------+  |        |  +------------+  |
       |  | DDoS       |<------------>| DDoS       |  |
       |  | Mitigation |C |        | S| Mitigation |  |
       |  | System     |  |        |  | System     |  |
       |  +------------+  |        |  +------------+  |
       +------------------+        +------------------+

           * C is for DOTS client functionality
           * S is for DOTS server functionality

       Figure 2: DDoS Mitigation between an Enterprise Network and Third
                 Party DDoS Mitigation Service Provider

[...]

In some cases the communication between the enterprise DOTS client and
the DOTS server of the DDoS Mitigation Service Provider may go through
the ITP carrying the DDoS attack, which would affect the
communication. On the other hand, the communication between the DOTS
client and DOTS server may take a path that is not undergoing a DDoS
attack.

</mglt>


> Section 3.3
>
>    In this use case, one or more DDoS telemetry systems or monitoring
>    devices monitor a network - typically an ISP network, an Enterprise
>    network, or a data center.  Upon detection of a DDoS attack, these
>    DDoS telemetry systems alert an orchestrator in charge of
>    coordinating the various DMS within the domain.  [...]
>
> nit: do we have a standard plural form for "DMS"?  (Is it just "DMS"?)
>
> <mglt>
Seems that DMS's is more appropriated. This has been changed.

Upon detection of a DDoS attack, these DDoS
telemetry systems alert an orchestrator in charge of coordinating the
various DMS's within the domain.

</mglt>


>    ITP.  DDoS Mitigation System selection and DDoS Mitigation technique
>    may depends on the type of DDoS attack.  In some case, a manual
>
> nit: "techniques" plural
>
> <mglt>
Corrected:

DDoS Mitigation System selection and DDoS Mitigation techniques may
depends on the type of DDoS attack.
</mglt>

   The communication between a network administrator and the
>    orchestrator is also performed using DOTS.  The network administrator
>    via its web interfaces implements a DOTS client, while the
>    Orchestrator implements a DOTS server.
>
> nit: as written, this is saying that the network administrator has a
> web interface.  I think "its" is supposed to refer to something else.
>
> <mglt>
What we are trying to say is that the network administrator sees its web
interface, and instruct the DOTS client from that interface. I have not
made any change to address that concern, as I do not clearly see what is
confusing.
</mglt>


> nit: Figure 4 lists "DDoS mitigation systems" in both the interprise and
> the ITP, but only the enterprise side has a "stack" of boxes to indicate
> there is more than one.
>
 <mglt>
Both have two DMS's now.


</mglt>

>
>    These systems are configured so that when an event or some
>    measurement indicators reach a predefined level to send DOTS
>    mitigation request to the orchestrator.  The DOTS mitigation request
>
> nit: the grammar here is a bit off; I think s/to send DOTS mitigation
> request/they send a DOTS mitigation request/ would fix it.
>
> <mglt>
Thanks. This has been corrected accordingly:

These systems are configured so that when an event or some measurement
indicators reach a predefined level they send a DOTS mitigation request
to the orchestrator.

</mglt>


>    Upon receipt of the DOTS mitigation request from the DDoS telemetry
>    system, the orchestrator responds with an acknowledgment, to avoid
>    retransmission of the request for mitigation.  The orchestrator may
>    begin collecting additional fined grain and specific information from
>
> nit: "fine-grained"
>
>    and provide an analysis of the event.  Eventually, the orchestrator
>    may ask additional information to the DDoS telemetry system, however,
>    the collection of these information is out of scope.
>
> nit: s/ask additional information to/ask for additional information
> from/
> nit: semicolon before "however" instead of comma
> nit: "this information"
>
> <mglt>
Thanks for the nits. All have been addressed in the text below:

The orchestrator may
begin collecting additional fine-grained and specific information from
various  DDoS telemetry systems in order to correlate the measurements
and provide an analysis of the event. Eventually, the orchestrator may
ask for additional information from the DDoS telemetry system; however,
the collection of this information is out of scope.

</mglt>


>    Upon receiving a request to mitigate a DDoS attack performed over a
>    target, the orchestrator, may evaluate the volumetry of the attack as
>
> nit: no comma after "the orchestrator"
>
>    well as the value that represent the target.  The orchestrator may
>
> nit: "the value that the target represents"
>
> <mglt>
The text has been corrected as follows:
Upon receiving a request to mitigate a DDoS attack performed over a
target, the orchestrator may evaluate the volumetry of the attack as
well as the value that the target represents.
</mglt>


>                           When DDoS Mitigation is requested, the status
>    indicates the DDoS Mitigation is starting while not effective.  The
>    DOTS client of the orchestrator will later be notified that the DDoS
>    Mitigation is effective.
>
> I'm not entirely sure what this last sentence is trying to say.
>
> <mglt>
Initially, I believe we wanted to distinguish between accepting the
mitigation and having the mitigation effective. I do not think that
necessary here as such details are provided in section 3.1.

The orchestrator requests a DDoS Mitigation to the selected
DDoS mitigation systems via its DOTS client, as described in Section
3.1.
</mglt>

>    Orchestration of the DDoS mitigation systems works similarly as
>    described in Section 3.1.  The orchestrator indicates with its status
>    whether the DDoS Mitigation is effective.
>
> Is this intended to specifically refer to the external (ITP) DMS?
>
> <mglt>
We did not see any differences between the DMS. I believe that the text
above clarify the DMS is the one selected by the orchestrator.

</mglt>


> Also, my understanding is that for this interaction the orchestrator is
> acting as a DOTS client, but the rest of the document only has status
> messages being generated by the DOTS server.  Am I confused?
>
>
<mglt>
The orchestrator got DOTS client and DOTS servers. I updated the section to
clarify which entity is involved in term of DOTS. I believe that clarifies
the concern.

</mglt>

> Section 4
>
> It feels incomplete to list three primary attacks but only discuss
> mitigations for two of them.  Perhaps "preconfigured mitigation steps to
> take on the loss of keepalive traffic can partially mitigate signal
> blocking, but in general it is impossible to comprehensively defend
> against an attacker that can selectively block any or all traffic".
>
> <mglt>
Just added this sentence.
</mglt>

> Section 6
>
> Is Med's name spelled correctly?
>
> <mglt>
Now it is spelt correctly! Thanks.
</mglt>

>
> Thanks,
>
> Ben
>
> _______________________________________________
> Dots mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/dots
>