[Ndn-interest] NFD-forwarding strategy retry

Sun Sep 30 07:13:23 PDT 2018

I have read this paper, maybe I will follow this work with different solutions.

Thanks for your helpful reminder. 

Shanshan

-----Original Messages-----
From:"David R. Oran" <daveoran at orandom.net>
Sent Time:2018-09-30 21:54:33 (Sunday)
To: Shanshan <shishanshan at cstnet.cn>
Cc: "Lixia Zhang" <lixia at cs.ucla.edu>, ndn-interest at lists.cs.ucla.edu
Subject: Re: [Ndn-interest] NFD-forwarding strategy retry

The paper https://conferences.sigcomm.org/acm-icn/2016/proceedings/p50-carofiglio.pdf
seems highly relevant to this discussion.

Have you read it?

On 30 Sep 2018, at 1:23, Shanshan wrote:

Dear Prof. Lixia Zhang,

Thanks for your reply.

1)given we are doing research here: could you elaborate the advantages from your planned approach over the current link layer reliability enhancement? 

The main advantage is a fine-granularity or qos-based loss detection and recovery: I have read the Hop-By-Hop Best Effort Link Layer Reliability in NDN (BELRP) and the NFD-developer-guide, and I am not fully sure if the sequence number and ack based solutions at the link layer can see the flow information. but as described in BELRP, if the BELRP is enabled, all packets require sequence, so I guess all packets require loss detection and retransmission (with the same max number of retransmission) without distinguish flow or service. However, packets loss detection and recovery have a trade-off between content retrieval delay and bandwidth consumption, so some delay sensitive or real-time applications require higher priority in packets loss detection and recovery, and NDN forwarding layer can easily achieve it, because every ndn node has PIT. 

Another advantage seems weak, and it is from a common assumption in the TCP/IP network: the wireless link is not completely reliable even relying on the link layer to cope with it, so reliable communication relying on high layer is necessary.

2)for point-point channels, the NDN-LP based solution seems working well, although this solution (so far) does not apply to multi-access channels. 

I am not sure I have fully understood the principle of NDN-LP especially about the granularity problem, and I will keep on reading the related papers, or could you recommend some related papers? I found that Link Service can work on top of several types of transport (in the NFD-developer-guide) but wireless/802.11 is absent, so are there any challenges in implementation of NDN over wireless different from IP over wireless, or is it the multi-access channels?

3)it is unclear to me why the FIB entry can only has one outgoing face (or this is just your assumption)

Yes, this is my assumption. The case is that producer (single-homed) connects to the edge router/BS/AP (via one hop) over a wireless channel. Some works can be done about timer-based retransmission on edge router/BS/AP. A more accurate timer can be estimated.

4)it does not seem hard to add such code into ndnSIM to try it out

Yes, it is. But I am not so sure if any existing forwarding strategy has this feature, or my planned approach has already overlap with them. The in-network retransmission seems to retry the Interest to a different alternative nexthop, and the goal seems to solve the multipath problem.

Best,

Shanshan

-----Original Messages-----
From:"Lixia Zhang" <lixia at cs.ucla.edu>
Sent Time:2018-09-30 01:04:52 (Sunday)
To: Shanshan <shishanshan at cstnet.cn>
Cc: "Teng Liang" <philoliang at email.arizona.edu>, ndn-interest at lists.cs.ucla.edu
Subject: Re: [Ndn-interest] NFD-forwarding strategy retry

On Sep 28, 2018, at 9:22 PM, Shanshan <shishanshan at cstnet.cn> wrote:

Hi Teng,

1)Why resend an Interest to the same outgoing face:

When a producer (single-homed) connects to (via one hop) an over a wireless channel, packets (Interest, Data, Nack) may get lost due to poor wireless link. I want to design a new forwarding strategy for the edge router/BS/AP for reliable communication at NDN fw layer rather than at link layer.

given we are doing research here: could you elaborate the advantages from your planned approach over the current link layer reliability enhancement?

for point-point channels, the NDN-LP based solution seems working well, although this solution (so far) does not apply to multi-access channels. 

In such case, the FIB entry only has one outgoing face.

it is unclear to me why the FIB entry can only has one outgoing face (or this is just your assumption)

2)and when:

In the above scenario, I want to design a new forwarding strategy for the edge router/BS/AP by setting a new timer (retransmission timer) for each PIT entry, and the retransmission timer is estimated by the RTT between each Interest and the corresponding Data belonging to the same flow. When the timer expires, the forwarding strategy retries this Interest to the same outgoing face. 

I also configure the Access Router Strategy and the ASF Strategy for the edge router/BS/AP in ndnSIM2.6, but neither implements Interest retransmission generated by the local node.

it does not seem hard to add such code into ndnSIM to try it out

-----Original Messages-----
From:"Teng Liang" <philoliang at email.arizona.edu>
Sent Time:2018-09-29 02:57:53 (Saturday)
To: Shanshan <shishanshan at cstnet.cn>, ndn-interest at lists.cs.ucla.edu
Cc:
Subject: Re: [Ndn-interest] NFD-forwarding strategy retry

Hi Shanshan,

Can you be more specific on the why and when you want to resend an Interest to the same outgoing face?

Teng

On 9/26/18 12:05 AM, Shanshan wrote:

hi,community

I read the latest version of the nfd-developer-guide and I am not sure if it is allowed to retry the same Interest (generated by the local node rather than the consumer) via the same outgoing face. 

Thank you!

Best,
Shanshan.

_______________________________________________
Ndn-interest mailing list
Ndn-interest at lists.cs.ucla.eduhttp://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

--
Shanshan Shi
PhD candidate
Computer Network Information Center,
Chinese Academy of Sciences
email: shishanshan at cstnet.cn

_______________________________________________
Ndn-interest mailing list
Ndn-interest at lists.cs.ucla.edu
http://www.lists.cs.ucla.edu/mailman/listinfo/ndn-interest

DaveO

--
Shanshan Shi
PhD candidate
Computer Network Information Center,
Chinese Academy of Sciences
email: shishanshan at cstnet.cn
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.lists.cs.ucla.edu/pipermail/ndn-interest/attachments/20180930/5fa7eccf/attachment-0001.html>