Project

General

Profile

Feature #5060

Feature #1624: Design and Implement Congestion Control

Congestion Detection: Monitor size of PIT vs. available memory

Added by Klaus Schneider 6 months ago. Updated 6 months ago.

Status:
New
Priority:
Normal
Assignee:
-
Category:
Tables
Target version:
-
Start date:
Due date:
% Done:

0%

Estimated time:

Description

The main idea here is to avoid overflowing the available memory for PIT entries, by proactively adding congestion marks once the PIT grows too large.

However, the exact algorithm is still work-in-progress. Feel free to discuss suggestions below.


Files

PIT-full_20170301.docx (24.2 KB) PIT-full_20170301.docx Junxiao Shi, 11/27/2019 06:28 PM

Related issues

Has duplicate NFD - Feature #1301: Shrink table size during memory stressNew

Actions
#1

Updated by Junxiao Shi 6 months ago

  • Has duplicate Feature #1301: Shrink table size during memory stress added
#2

Updated by Junxiao Shi 6 months ago

I thought about this long ago. See notes in attachment.
The main design challenge is fairness. Using congestion marks some alleviates this challenge, if everybody is cooperating.

On implementation side: do you want to enforce a fixed PIT capacity limit, or read system RAM usage?

#3

Updated by Davide Pesavento 6 months ago

  • Category set to Tables
  • Start date deleted (11/27/2019)
#4

Updated by Klaus Schneider 6 months ago

Junxiao Shi wrote:

I thought about this long ago. See notes in attachment.
The main design challenge is fairness. Using congestion marks some alleviates this challenge, if everybody is cooperating.

Yeah, I wouldn't worry too much about it, and just leave fairness to the endpoints.

On implementation side: do you want to enforce a fixed PIT capacity limit, or read system RAM usage?

Not sure yet. I assume there has to be some sort of best practice how to deal with limited memory?

Fixed PIT limit is likely easier to implement.

Also available in: Atom PDF