3 ms·
Personally I very much like the idea of having multiple pipes for different classes of traffic, as long as I (and/or the apps I'm running) get to choose what go
by musicale 6y ago
Personally I very much like the idea of having multiple pipes for different classes of traffic, as long as I (and/or the apps I'm running) get to choose what goes where. Presumably that would be permitted under the California law?
For example, gaming traffic and audio conferencing/telephony need low latency, which requires short queues (to minimize queuing delay) or no queues at all (e.g. cut-through routing.) However, short queues and cut-through routing break if the incoming data exceeds channel capacity. Typical solutions are separate queues/queue slicing and/or rate-limiting/throttling, and I'm completely OK with that because it enables low-latency, low-bandwidth applications like gaming and audio.
What I really don't want is a buffer bloat situation where everything goes into the same giant queue and gaming/audio traffic has to wait behind Netflix, software updates, cloud backups, etc..
There's a reason we have express checkout lanes at grocery stores, after all, as they tend to reduce queuing delay for people with fewer items.
- rhexs 6y agoI'm curious -- are there any examples of what ping time would be from SF to NY if something specific like this existed? How much of current (additional) ping is due to queueing vs speed of light in copper? Or would this just make it more reliable -- instead of jumping around a bit, a consistent connection with reliable, predictable responses?