[ndnSIM] placement policy classes.

[Alex Afanasyev]

Hi Andriana,

>From what you're describing, the placement policy is still very related to cache replacement policy.  At least you can consider your problem in terms of replacement. That is, until cache space is available, you don't need to do anything else. As soon as there is a new item to be cached,  the policy (I intentionally omitted word "replacement", since you can do more than just replacement with these policies) need to determine what to do with a new item.

If you are maintaining additional metrics, probabilities, information about neighbors (though I'm not sure what it means for individually cached data packet), you can ignore caching of a new item, evict some item from the cache and cache new item, invoke specialized protocol to move some data packets to other caches (not sure how, but it is technically possible).

Is it what you are asking?

---
Alex

On May 23, 2013, at 7:07 AM, ioannoa at scss.tcd.ie wrote:

> Hello,
> 
> I want to create some
> classes implementing different placement policies. As far as i've seen
> there are no placement policies implemented. That would be something
> similar to the replacement classes that you have in the utilities file i
> suspect.
> 
> Could you please provide me with any information possible on this? I can
> send you over the policies for further information if necessary. I also
> thought, i may try to develop them and send it over to check them. To be
> honest I am not getting much understanding of C++ code, so any help would
> be appreciated.
> 
> Best regards,
> Andriana.
> 
> 
> 
>> Can you tell a little bit about what do you mean by "placement" policy?
> ---
>> Alex.
> 
> 
> Hello,
> 
> yeah, sure.
> 
> So, there are some 'placement' policies in ICN literature - and by this do
> not only mean CCN. CCN follows the placement policy of caching each packet
> that passes by a node-router - using also a replacement policy such as
> LRU, LFU etc when the content store is full.
> 
> However, in other ICN proposals different policies may be used, e.g. based
> on a probability, based on graph-based metrics, e.g. highest number of
> neighbors etc. Not, too sure if that would be part of the open source
> though as you are supposed to implement CCN at first place.
> 
> Did I make any sense? ;)
> 
> Andriana.
> 
> 
> 
> 
> 
> 
> 
> 
> 
> 
> _______________________________________________
> ndnSIM mailing list
> ndnSIM at lists.cs.ucla.edu
> http://www.lists.cs.ucla.edu/mailman/listinfo/ndnsim