My very basic understanding is that over limits is about packets that can not be sent immediately when they arrive because that would cause the send rate to go over the limiting send rate. It's normal to have some over limits packets when shaping.
The fact that you have only 4 drops indicates that the overlimits is do to burstyness not sustained duration of overlimits
I think you are right, drop-mode is certainly not the right way to describe that.... and cobalt/blue is driven by individual queues staying full which is related to overlimits but not very strictly, as you indicate bursty traffic can cause overlimit count to build up without resulting in loner term queues developing (which would result i drops/marks).
The technician came today and found some thousand FEC, 2 CRC and some EC or ES which are very revealing regarding the line from the central to the input box.
Solution: either replace the entire wiring or use another pair of wires
Concerning my problem of crosstalk, we can remove or greatly reduce this phenomenon by using the vectoring ... but the card to the central is not provided ... which is unfortunate because it stabilizes the line enormously and can triple the bandwidth if we are near the dslam.
These DSL counters really are only interpretable if the links uptime is known....
FEC, or (successful) forward error corrections, and often not even individual correction counts but a count of individual seconds with >= 1 FEC event. These are pretty benign as these are checked and corrected as part of normal receiver processing.
CRCs, or cyclic redundancy check errors, these are not benign as each of those denotes a lost packet (or a second with lost packet(s)). But low counts can easily be ignored....
No Idea right now what EC might be...
ES however counts the errored seconds, that 8s seconds with one or more code violations like CRCs. If the count is low that can IMHO also be ignored...
About vectoring, it is a bit of a mixed blessing, yes it can help but does so by expecting all modems to behave well. 
P2MP is a widespread network architecture
internationally and nationally
P2Mpi
Mainly P2MP
Some territories in P2P
Mainly
P2P
The P2MP architecture is recognized and used on the international
the international scene - many telecommunication operators use it
operators use it: for example A1 in Austria, Deutsche Telekom in
Austria, Deutsche Telekom in Germany, KPN in the Netherlands
Orange in France and TIM in Italy.
The P2MP architecture has become established against P2P
for FTTH networks on an international scale. By
Consequently, future innovations and developments are
future developments are strongly focused on this technology…
Which of the 2 favors a better latency ???
P2P or P2MP are mainly descriptions on the "wiring", for fiber either there is a dedicated fiber between the CPE (fiber-modem) and the next active element (point to point, often used for active optical networks, AON), or there is a shared fiber segment between CPEs and first active element (point to multi-point, used for passive optical networks, PON). PON tends to be cheaper (less fibers to lay, less space required in the CO, less active ports required and hence less energy used) than AON and it keeps stricter control in the hands of the ISP that built the PON plant. With P2P connections one can still add a passive splitter at the CO and operate these as PON (often still attractive). So the whole economics and control issues basically make PON more likely.
PON requires some arbitration mechanisms between the different senders (request-grant mechanism, similar to cable/docsis, albeit with a considerably shorter delay, 0.125ms instead of >= 2 ms) which will introduce some (variable) delay, so IMHO AON has the potential of having lower latency here, BUT we are talking probably less than a single millisecond here, so potentially an issue that might show up for intra-rack/intra-datacenter traffic, but nothing to worry about for normal internet traffic.
Thank you for its precision, the total fiber is still very far for me, at best Gfast for the end of July ... in the meantime I have vdsl2 not at all responsive ... I am waiting for the replacement of copper cable (200m) then I guess I'll change ISP, I doubt that it changes my problem of latency ... to see .
Hello to all, just to keep you informed. My ISP is in investigation mode for my latency, they recently asked me for the IP address of problematic game servers. I am unfortunately forced to leave my ISP router on for analysis purposes ( I guess I am monitoring) I have added my OpenWrt belkin router with the qosify script but no way to improve responsiveness ....Waiting for their investigations to be completed. If there is no way out of this latency I will be forced to change ISP.
What do you think of this DSCP class table issued by cisco? Does it seem coherent and functional?
Sounds like a good solution.
It's been pretty clear for a while now that it's not an OpenWrt issue.
The cancellation period for the standard subscription is 60 days, but if a problem is not solved, as is the case for me, then the period is 7 days. If changing ISP changes absolutely nothing, I still have the advantage of paying less for the same services!
This is a nice/concise summary table from the referenced RFCs, but this is only functional if the bottleneck actually is configured to treat packets differently according to these DSCP values. DSCPs are really just numerical values in a 6bit field of the IPv4/6 header, any meaning is only assigned if a network drives a differential per hop behaviour (PHB) based on these values.
Intermediate nodes in the network are unlikely to honor DSCPs you set with the expected differential behaviour unless you negotiate a service level agreement (SLA) and then theoretically you would need to do this with all networks between your network and your target servers.
hello, I finally changed isp, and guess what? I have exactly the same latency, delay, heaviness, no reactivity no consistency.. now I think I have found the problem but not the solution.. on the COD game I receive 120 tx rate and I send rx rate 60… it sends me missing 60… then on fifa ke receive Tx rate 30 and I read rx rate 30.. when we should be playing at 60hz.. if I miss 60 frames per second I have to be late and if I play at 30 and my opponent is playing at 60 I miss 30 hz .. and what all this explains my latency and if so, what is the solution?
Yes, it's the game or the servers. The solution is either play on faster tick rate servers if there are any... or play a different game.
thank you Daniel, in other words there is no solution? no because this is the case for all my games .. is it possible that my ISP is slowing down the frame rate ??? or is it possible that this is a way for game servers to compensate for connections? that said it may be normal that I have the same latency symptoms because he rents the line to my old ISP ..
Well, your DSL modem should allow you to get information about the actual DSL link, but anything after that to the game servers is harder to diagnose, starting with the uplink of the DSLAM.
Lately I am trying to convince people to use the following tool to get some basic DSL statistics and plots:
Maybe you can try to use that (assuming it works with your modem), use the GUI and let it run for a few hours if possible as it collects maximum and minimum values for the SNR margin, which can be diagnostic of external noise ingress.
Other than that, which ISP did you end up with?
thank you I will read this and try to understand .. I opt for Init7 as ISP .. I have the advantage of putting the router I want, I now use a pppoe connection.
I have a mikrotik RB5009 router with sfp
| Architecture | ARM 64bit |
|---|---|
| CPU | 88F7040 |
| CPU core count | 4 |
| CPU nominal frequency | 350-1400 (auto) MHz |
| RouterOS license | 5 |
| Operating System | RouterOS (v7 only) |
| Size of RAM | 1 GB |
| Storage size | 1 |
This 2 graffics from netduma is from what game.
Normaly for cod the server is from choppa and vult not from googleusercontent.com
I would have to test with netduma
but I'm on rb5009 for the moment
you are doing a test on fifa and the other cod is it ?
yes strangely the fifa servers which were mainly amazon (aws) have become googleusercontent.. is this temporary I don't know.. I have the same tickrate problem on COD except that it goes from 120 to 60hz then 120 to 60 hz everything along the game
