[Ndn-interest] Question regarding consistency

Muhammad Hosain Abdollahi Sabet M.AbdollahiSabet at mail.sbu.ac.ir
Mon Aug 24 22:59:45 PDT 2015


Aniesh,
You mean most of the time Tw(as ChronoSync named it) expired and your nodes did processing of a corresponding sync interest, so missed some others-generated data?

Sabet


-----Original Message-----
From: Ndn-interest on behalf of Lixia Zhang
Sent: Tue 8/25/2015 9:52 AM
To: aniesh chawla
Cc: ndn-interest
Subject: Re: [Ndn-interest] Question regarding consistency
 

> On Aug 24, 2015, at 10:00 PM, aniesh chawla <chawla.aniesh at gmail.com> wrote:
> 
> Thanks everyone.
> 
> Lixia,
> you were right, I was looking for the third point.
> 
> Wentao,
> Thanks for the clarification about NDNFS
> 
> So If I run ChronoSync, it will provide me with eventual consistency but there are timers associated with them which might make things bit difficult and complex. Do you think that is timers(as in ChornoSync) is the right way to go forward or was it more like Proof Of Concept to showcase that synchronization is possible? Personally I don't like timers which are associated with network (in this case propagation delay), I had really bad experience when I did beacon synchronization which was almost like the one mentioned in ChronoSync in my earlier project and every time we end up changing the timers.
> 
> Regards
> Aniesh

I am not sure what/which ChronoSync "timers" you are concerned about ...
Sync is not just "possible", it's an importance piece in the architecture.

Would be interested in hearing more about your lessons from your beacon sync. 

The specific design of ChronoSync, as described in ICNP 2013 paper, has its limitations; a revision has been worked on, hope to make it solid enough to share soon.

Lixia

> On Tue, Aug 25, 2015 at 9:22 AM, Wentao Shang <wentaoshang at gmail.com <mailto:wentaoshang at gmail.com>> wrote:
> 
> > On Aug 24, 2015, at 6:12 AM, aniesh chawla <chawla.aniesh at gmail.com <mailto:chawla.aniesh at gmail.com>> wrote:
> >
> > Hello Everyone,
> >
> > I have a very basic doubt which I could not resolve while reading documentation of NDN. I would like to also take this as research topic if my doubt is valid.
> >
> > The question is:
> > How is the consistency of data maintained in the network? Since, NDN is "pull" based networking and lets say there is change in file version at one of the node and not in other nodes. I can see the this can be achieved through NDNFS, but still the other nodes will not have change in version of the file. So, how will the new node who is requesting for that file will get to know which version to ask for?
> 
> Hi Aniesh,
> 
> As Lixia said in her reply, NDN is a layer-3 protocol. While it moves some of the application layer functionalities into the network layer, it cannot solve every problem the application faces.
> 
> NDNFS is a piece of software that demonstrates the idea of integrating application data unit (i.e., file blocks) with network data unit (i.e., NDN data packets). It was originally designed to be a local file system without any remote replication. Therefore the data consistency issue was conveniently ignored.
> 
> If you need synchronization across multiple copies of the data, you may run your application on top of ChronoSync, which gives you some form of eventual consistency.
> 
> Best,
> Wentao
> 
> > If we use NDNS to resolve every such query and we use time stamp as version then also it should be propagated throughout the network? I also see that Chrono-sync does something like this but it has waiting timer for propagation delays. So either NDNS is talking to every node in the whole network or some other way.
> >
> > Is my understanding correct or am I missing something here?
> >
> > Thanks a lot
> >
> > Regards
> > Aniesh
> > _______________________________________________
> > Ndn-interest mailing list
> > Ndn-interest at lists.cs.ucla.edu <mailto:Ndn-interest at lists.cs.ucla.edu>
> > http://www.lists.cs.ucla.edu/mailman/listinfo/ndn-interest <http://www.lists.cs.ucla.edu/mailman/listinfo/ndn-interest>
> 
> 
> _______________________________________________
> Ndn-interest mailing list
> Ndn-interest at lists.cs.ucla.edu
> http://www.lists.cs.ucla.edu/mailman/listinfo/ndn-interest


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.lists.cs.ucla.edu/pipermail/ndn-interest/attachments/20150825/b6a39882/attachment.html>


More information about the Ndn-interest mailing list