[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [RRG] mapping "issues" list from this morning



Hi Dimitri.

On 3/15/08 3:35 PM, dimitri papadimitriou allegedly wrote:
hi - a couple of comments/suggestions here below:

"- Whether there will be any delay at all (might even get there
  faster)."

-> differential delay would better translate this issue

Good point. That's what we try to use when we talk about it anyway. I'll change the section title.

-> also consider mapping entries change/variation for ongoing traffic
flows and the side effect with corrupted entries

To paraphrase: the possibility of additional delay/drop for packets in established flows immediately after mapping entries change -- is that right?

Unintentional changes are covered under "Failure Modes". I'll add a line under "churn" for intentional changes.

(I don't think "delay" should be a major heading of its own, because delay is a result, not an event, but I don't know how to organize things yet so that the "delay" section goes away. When someone actually tries to use this list, we'll know how to organize it better.)

"- Number of packets that might be delayed, dropped."

-> add "effect on e.g. short elastic and long bulk flows"

"- Effect on transport layer, session."

-> for the transport layer (e.g. flow and congestion control)

Done

in "Push, Poll, Hybrids"

-> distinguish communication between MP and communication between TRs and MP

Is MP a Mapping Point? Do you mean the source of mapping information for the tunnel routers? Can you expand on this one a little?

-> in-band/user-driven trigger impact for mapping resolution (more than
longest match-prefix lookup as performed today)

add in "mapping points"

-> distribution of the mapping points vs modes (push, pull, hybrid)

-> impact of mapping point failure (for LISP 2 and 3 for inst. a there
is no fate sharing this would result in serious impact)

add issue of evolvability

-> EID allocation and structure (noticing that a 1:1 mapping does not
resolve any routing system scalability issue)

-> mapping system scalability and complexity

I can't find any of this in what I sent out. Could you send me what you are quoting from?

editorials

"Push, Poll, Hybrids"

-> replace by "Push, Pull, Hybrids"

Done.

I'm attaching the updates I've done, and a diff.

Scott

Support for other network needs
-------------------------------

- Mobility.
  - Impact on preferences for mobility approaches
  - Does it meet mobility requirements
- Emergency services.


Differential Delay and/or Drop of initial packets
-------------------------------------------------

- Whether there will be any delay at all (might even get there
  faster).
- Number of packets that might be delayed, dropped.
- Between sites and/or hosts.
- Trade off determinism of delay/drop versus agility & light weight.
  Analyze based on how bad it would be.
- Effect on transport/session layers (e.g. congestion control).
- Effect on short elastic and long bulk flows.
- Ultimately only end user experience is visible.
- Possibility of difference classes of service?


Push, Pull, Hybrids
-------------------

- Possibly proactive, dynamic adjustment of behavior.
  - By src/dst (site)
  - Responsiveness
  - Other parameters?


Security considerations
-----------------------

- Interactions with drop/delay.  (Eliot Lear -- needs clarification
  please).
- Authorization: Don't take requestors down garden path.  Don't give
  them information they aren't authorized to actually use.
- Security of mapping system itself: updates authenticated .
- Authentication of map replies.
- Interactions between security of mapping system and security of core
  routing.


Granularity
-----------

- What level of specificity?
- Ability to operate at multiple levels of granularity simultaneously.

Churn
-----

- How system responds with change in rate of intentional changes.
- Effect of changes to mappings on established flows.

Mapping points
--------------

- Flexibility in mapping points: end system, domain border, in
  between.  
- Placement of mapping points: those that own the mappings, those that
  use them, or in a third party
- Who is authoritative for the mapping.  Owner, holder, user.


Interaction between rate, state, and delay
--------------------------------------

- Interaction between rate, state, and delay.


Deployability (Incremental)
---------------------------

- Interworking with existing deployments.
- How soon are benefits seen?
- Who gains.  Is it the one doing the deployment?
- Does a deployment require more than one party, coordinated?
- Easing v4=>v6 is not an issue.


Flexibility
-----------

- Possibility for hierarchy
- Ability to express preferences (policy)


Failure modes
-------------

- How does mapping system adapt to unintentional changes in reachability (if at
  all).
- Dependence on something not on the data path (-> uncontrolled  weaknesses).
- Resilience of mapping system.


*** /Users/sbrim/working/routing/mappingCriteria_080312.txt	Wed Mar 12 10:10:15 2008
--- /Users/sbrim/working/routing/mappingCriteria.txt	Tue Mar 18 16:38:02 2008
***************
*** 7,14 ****
  - Emergency services.
  
  
! Delay and/or Drop of initial packets
! ------------------------------------
  
  - Whether there will be any delay at all (might even get there
    faster).
--- 7,14 ----
  - Emergency services.
  
  
! Differential Delay and/or Drop of initial packets
! -------------------------------------------------
  
  - Whether there will be any delay at all (might even get there
    faster).
***************
*** 16,27 ****
  - Between sites and/or hosts.
  - Trade off determinism of delay/drop versus agility & light weight.
    Analyze based on how bad it would be.
! - Effect on transport layer, session.
  - Ultimately only end user experience is visible.
  - Possibility of difference classes of service?
  
  
! Push, Poll, Hybrids
  -------------------
  
  - Possibly proactive, dynamic adjustment of behavior.
--- 16,28 ----
  - Between sites and/or hosts.
  - Trade off determinism of delay/drop versus agility & light weight.
    Analyze based on how bad it would be.
! - Effect on transport/session layers (e.g. congestion control).
! - Effect on short elastic and long bulk flows.
  - Ultimately only end user experience is visible.
  - Possibility of difference classes of service?
  
  
! Push, Pull, Hybrids
  -------------------
  
  - Possibly proactive, dynamic adjustment of behavior.
***************
*** 52,59 ****
  Churn
  -----
  
! - How system responds with rate of intentional changes.
! 
  
  Mapping points
  --------------
--- 53,60 ----
  Churn
  -----
  
! - How system responds with change in rate of intentional changes.
! - Effect of changes to mappings on established flows.
  
  Mapping points
  --------------

Diff finished.  Tue Mar 18 16:38:52 2008