I don't think its just number of channels, e.g. SH2 is max 8 channels almost always, but I've seen variations about 2-3 times the base latency.
I think it depends on how good your cabling is and how many other people on the network with you - remember, you sharing cabling bandwidth, and even if its enough throughput, latency still will be affected. Unlike DSL which puts everyone on separate port&cable run.
E.g. this is my BQM (SH2 at maximum 8 channels), I have cabinet right outside my house and I am pretty sure there only 2-3 other persons on same cable:
Most of SH2 graphs I've seen are significantly worse, even without apparent congestion, just at higher avg. max latencies (~60ms). But above graph already at 8 max channels, so these people can't get it any worse because of more channels - hence their network neighbourhood is only possible culprit.
@RidingTheFlow wrote:I don't think its just number of channels, e.g. SH2 is max 8 channels almost always, but I've seen variations about 2-3 times the base latency.
I think it depends on how good your cabling is and how many other people on the network with you - remember, you sharing cabling bandwidth, and even if its enough throughput, latency still will be affected. Unlike DSL which puts everyone on separate port&cable run.
E.g. this is my BQM (SH2 at maximum 8 channels), I have cabinet right outside my house and I am pretty sure there only 2-3 other persons on same cable:
Most of SH2 graphs I've seen are significantly worse, even without apparent congestion, just at higher avg. max latencies (~60ms). But above graph already at 8 max channels, so these people can't get it any worse because of more channels - hence their network neighbourhood is only possible culprit.
That is why I said "there are other variables of course", but this topic is about the puma 6 issue. There are plenty of reasons why 2 BQM's will differ will look in general, but none of which I know to have a direct relation with the Puma 6 issue like the number of channels.
Network neighbourhood (Are we back in NT 4.0 days 
) assuming you mean network hub to cab/cab to CMTS is just one reason.
@Guybrush85 wrote:
That is why I said "there are other variables of course", but this topic is about the puma 6 issue. There are plenty of reasons why 2 BQM's will differ will look in general, but none of which I know to have a direct relation with the Puma 6 issue like the number of channels.Network neighbourhood (Are we back in NT 4.0 days
I don't argue that having more channels on P6 generally makes it worse, but its not always consistent (and above graph with 24 channels proves it).
The factors I've meant:
- Your cable plant topology and quality (e.g. if any of your neighbors around have lose cut of cable ends without terminators causing microreflections and "false collisions")
- How above affects DOCSIS negotiation and resulting channel number assignment, modulation type and different bonding timings as a result
- Your CMTS model obviously, which also will change things quite drastically. CMTS can cause latency spikes as well as modem does
Btw actual "yellow spike" issue may be as stupid as superhub having ICMP on low priority (read here - https://www.thinkbroadband.com/faq/broadband-quality-monitor, "My graph has lots of very regularly spaced yellow spikes on it! What is wrong with my connection?").
However, as Datalink pointed out there are two issues here - icmp spikes (which may be easy fix), which BQM measures, and UDP latency spikes under load - which BQM will not measure.
@RidingTheFlow wrote:
@Guybrush85 wrote:
That is why I said "there are other variables of course", but this topic is about the puma 6 issue. There are plenty of reasons why 2 BQM's will differ will look in general, but none of which I know to have a direct relation with the Puma 6 issue like the number of channels.Network neighbourhood (Are we back in NT 4.0 days
I don't argue that having more channels on P6 generally makes it worse, but its not always consistent (and above graph with 24 channels proves it).
The factors I've meant:
- Your cable plant topology and quality (e.g. if any of your neighbors around have lose cut of cable ends without terminators causing microreflections and "false collisions")
- How above affects DOCSIS negotiation and resulting channel number assignment, modulation type and different bonding timings as a result
- Your CMTS model obviously, which also will change things quite drastically. CMTS can cause latency spikes as well as modem does
Btw actual "yellow spike" issue may be as stupid as superhub having ICMP on low priority (read here - https://www.thinkbroadband.com/faq/broadband-quality-monitor, "My graph has lots of very regularly spaced yellow spikes on it! What is wrong with my connection?").
However, as Datalink pointed out there are two issues here - icmp spikes (which may be easy fix), which BQM measures, and UDP latency spikes under load - which BQM will not measure.
That graph is an odd one, hence my inquest into futher information. My BQM was pretty much like for like with with my 2ac and similar to yours above, this only worsened when my channels increased. There's lot of other work in the area, none of which has caused any noticable difference both BQM or other testing. So again, we can talk about all the variable that may cause a BQM to differ, but unless you can correlate one to the Puma 6 issue, I don't understand the relavance?
