I haven't been able to find information about what the "Number of bytes" column for Classification Rules in the luci-app-qos module is for. Can anyone enlighten me? Thanks.
I don't really know for sure. But given that what you are able to adjust in this line I would assume that the value is an assured (min.) number of bytes for the target/ports. So you can fine tune certain applications to your needs.
Use SQM (luci-app-sqm), Luke. Simple and it actually works.
qos has per-host prioritization, which was exactly what I needed, and it worked just fine.
When I set it, connbytes
appeared:
Rules
Each
classify
section defines one group of packets and which target (i.e. bucket) this group belongs to. All the packets share the bucket specified.
Name Type Required Default Description connbytes
int no (none) Packets matching this, belong to the bucket defined in target
~ From: https://openwrt.org/docs/guide-user/network/traffic-shaping/traffic_shaping#rules
From reading this I would phrase it as: "a set of packets matching this totaling greater-than X bytes transferred".
Also see: http://luxik.cdi.cz/~devik/connbytes/
Ok, so the shaping isn't applied to a particular connection until it reaches a specified minimum byte threshold. Got it. Thanks.
If your problem is solved, please consider marking this topic as [Solved]. See How to mark a topic as [Solved] for a short how-to.
If that is true, this rule can be used as a heuristic to detect bulk transfer flows, to allow to de-prioritize them to give other potentially interactive flows lower latency.
SQM's qdiscs fq_codel and cake solve the interactivity issue slightky differently. Firstly, they default to flow-fair queueing, which will put al flows on equal footing and secondly, they will also give flows without a backlog (that is without packets already in the queue) a slight boost over flows with backlog. Bulk transfers tend to accumulate some queue, while interactive flows tend not to. That said, there are still people who manually combine a number-of-bytes type rule to dscp-mark bulk flows in combination with layer_cake, where one can, in addition to the default modest sparseness boosting, also steer packets into multiple different priority tiers....
I didn't have a "problem" that needed a "solution". I simply asked a question.
Yes, but if you are happy with the answer to the question, marking the best answer as solution, will help others that later find this thread to get up to speed quickly.