[Ndn-lib] PSync PartialProducer on multiple nodes

Lan Wang (lanwang) lanwang at memphis.edu
Wed Jan 30 06:47:49 PST 2019



On Jan 29, 2019, at 5:44 PM, Junxiao Shi <shijunxiao at email.arizona.edu<mailto:shijunxiao at email.arizona.edu>> wrote:

Hi Lan

Each producer has one UserNode in the FullSync group. If two producers are serving the same dataset, they should participate FullSync under the same UserNode. Consumers do not have UserNodes in the FullSync group.
I am not sure if we allow pure consumer nodes (no UserPrefix) in FullSync mode.

FullProducer constructor has a userPrefix parameter for "the prefix of the first user in the group", but then the application could remove this UserNode via FullProducer::removeUserNode.
Doxygen does not disallow it.

I didn’t know this removal functionality exists, but even if the userNode is removed locally others who have received it before will add it back through sync.  There is no way to do removal easily in group sync right now until we have a good security mechanism (to authenticate who can delete what otherwise anyone can delete anything).

Ashlesh: this removal functionality is desirable, but as I said above it is probably not working as you intended.  You will need to investigate it (and create a redmine issue if it doesn’t work).

One potential problem is what to do when multiple distinct producers have the same PartialSync UserNode.

If they have different data, they should not have the same user prefix.  If they do have the same data (e.g., they replicate each other’s data), then PartialSync can handle these producers having the same user prefix.

These PartialSync UserNodes are supposed to have the same Data, but library must have reasonable fallback on application's mistakes.
It's not up to PartialSync protocol to deal with the two producers with the same user prefix, because they are operating under different syncPrefixes.

If two producers have different sync prefixes, they don’t interfere with each other even if they have the same user prefix.

Lan

Yours, Junxiao

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.lists.cs.ucla.edu/pipermail/ndn-lib/attachments/20190130/b65f6dae/attachment.html>


More information about the Ndn-lib mailing list