[Nfd-dev] Why don't additional routes impact performance?

[Lan Wang (lanwang)]

Jeff,

The behavior you want can be achieved by the ncc strategy, but we have identified other undesirable problems with this strategy (an interaction of the sequential probing this strategy does and the duplicate suppression implemented in NFD).  See this nfd issue: http://redmine.named-data.net/issues/2592

Unfortunately there're no alliterative other than ncc and best-route in NFD right now.  Vince Lehman in my lab is working on implementing the GreenYedRed strategy (proposed in http://named-data.net/publications/comcom-stateful-forwarding/ and refined in http://named-data.net/publications/role_of_routing_ndn/), but it is not ready yet.

Lan

On Mar 29, 2015, at 12:03 PM, "Burke, Jeff" <jburke at remap.UCLA.EDU<mailto:jburke at remap.UCLA.EDU>>
 wrote:

Hi Junxiao,

Ok, that makes sense.  This was sort of an extreme case, though. What about for two hosts likely to have "equal" cost (e.g., UCLA and UA, for REMAP) that have different performance in practice for a given subnamespace that is not known a priori?

Jeff

From: Junxiao Shi <shijunxiao at email.arizona.edu<mailto:shijunxiao at email.arizona.edu>>
Date: Sun, 29 Mar 2015 09:42:35 -0700
To: Jeff Burke <jburke at remap.ucla.edu<mailto:jburke at remap.ucla.edu>>
Cc: "nfd-dev at lists.cs.ucla.edu<mailto:nfd-dev at lists.cs.ucla.edu>" <nfd-dev at lists.cs.ucla.edu<mailto:nfd-dev at lists.cs.ucla.edu>>
Subject: Re: [Nfd-dev] Why don't additional routes impact performance?

Hi Jeff

With default configuration, you are using best-route v3 strategy.
Therefore, this behavior is by design.

Please try: register BASEL route with higher cost than REMAP or CAIDA route.
When a route with lower cost is inserted, traffic shall be switched immediately to the new route.

Yours, Junxiao

On Sun, Mar 29, 2015 at 9:21 AM, Burke, Jeff <jburke at remap.ucla.edu<mailto:jburke at remap.ucla.edu>> wrote:

Hi,

I am running some tests with ndncon and am curious about the following:

Steps to reproduce:

- Start ndncon publisher with route through REMAP node.
- On consumer node,
- Restart nfd
- Use nfdc to register udp route for / through university of basel
- Start consumer and observe performance (audio dropouts due to jitter buffer issues)
- While ndncon is running, use nfdc to register udp route for / through REMAP, CAIDA, or another closer hub
- No change in performance at the consumer and no change in traffic flow on ndnmap... why?
- Quit ndncon and restart
- Still no change

This behavior is not intuitive.  I would expect that by default NFD would be issuing interests to both faces, and/or that it would be probing performance new faces become available for a given route.    Does a different strategy need to be used to get this behavior?

On the consumer, I am running the latest version of NFD cloned from github this morning, with a default configuration (I think).


Thanks!
Jeff


_______________________________________________
Nfd-dev mailing list
Nfd-dev at lists.cs.ucla.edu<mailto:Nfd-dev at lists.cs.ucla.edu>
http://www.lists.cs.ucla.edu/mailman/listinfo/nfd-dev


_______________________________________________
Nfd-dev mailing list
Nfd-dev at lists.cs.ucla.edu<mailto:Nfd-dev at lists.cs.ucla.edu>
http://www.lists.cs.ucla.edu/mailman/listinfo/nfd-dev

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.lists.cs.ucla.edu/pipermail/nfd-dev/attachments/20150329/f84c1656/attachment.html>